JP2003108483A - Supporting system for generating mail, method of controlling supporting system for generating mail, system for distributing mail, control program, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily generate and transmit an electronic mail beneficial to both a mail receiver and a mail transmitter, by conducting individualization processing fit to the receiver of the electronic mail, while reducing time and labor in a transmitter side and a load in a facility such as a server therein. SOLUTION: A subject analyzing engine 23 analyzes the mail text of a mail document using natural language processing, an individualization engine 24 controls a standardization engine 26 and a word selection engine 28, and edits the mail text using the natural language processing while corresponding to the each mail receiver, based on personal information corresponding to the each mail receiver, preliminarily set personal information and an analyzed result of the mail text, so as to generate the individualized mail text.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mail generation support device, a mail generation support device control method, a mail delivery system, a control program, and a recording medium, and in particular, edits a direct mail according to the taste of a mail recipient. It relates to technology for personalization and distribution.

[0002]

2. Description of the Related Art In the electronic mail relay device described in Japanese Patent Laid-Open No. 2000-270013, a technology for converting an electronic mail to reflect the request and preference of the recipient and posting an advertisement in which the user is interested. Is disclosed. The e-mail relay device performs a personalization process on an individual e-mail, such as a method of inserting an advertisement and a kanji-kana conversion method, according to the taste of each individual. However, what kind of personalization processing is performed is set by the individual user. In addition, the maintenance of the actual personalization process was supposed to be performed by the mail sending side.
Further, as a specific personalization processing method, only addition of information, paraphrase of information, and deletion of information have been performed.

Also, as a network system for distributing introduction information created by an information provider, "SION" proposed by NTT Mirai Net Laboratories is known. This network system is constructed as a collation network that can acquire introduction information and make announcements when the user and the information provider do not know each other's existence. This allows the user to select his / her interests and attributes (gender,
It has become possible to identify information that matches the age group, occupation, etc.) and environment, and to identify other users who are suitable for providing their own information by the information provider who is the user.

[0004]

However, in the above-mentioned conventional technique, when the e-mail or information to be delivered to an individual is to be varied according to the taste of the individual, a plurality of e-mail texts or a plurality of e-mail texts are used. It was necessary to prepare the above information on the mail transmitting side or the information distributing side. Therefore, for example, when trying to send a direct mail, it is not realistic for the sender side to prepare the personalized e-mail for each mail receiver, and it is not always meaningful for the mail receiver side. There was a problem that information could not be obtained, and that the sender of the email could not achieve the desired advertising effect. Therefore, an object of the present invention is to reduce the burden on the sender side and the load on equipment such as a server, while performing personalization processing according to the recipient of each e-mail, thereby An object is to provide a mail generation support device, a mail generation support device control method, a mail distribution system, a control program, and a recording medium that can easily generate and distribute an electronic mail useful for both parties.

[0005]

[Means for Solving the Problems] In order to solve the above problems, in a mail generation support device that supports generation of mails to be delivered to a plurality of mail recipients, the mail body of a mail manuscript is analyzed using natural language processing. An analysis unit, personal information corresponding to each of the mail recipients, preset personalization information, and a natural language process for the mail text corresponding to each of the mail recipients based on the result of the analysis. And an editing unit for generating a personalized mail text by editing. According to the above configuration, the analysis unit analyzes the mail text of the mail manuscript using natural language processing. The editing unit edits the mail text using natural language processing for each mail recipient based on the personal information corresponding to each mail recipient, preset personalization information, and the analysis result of the mail text. Then, the personalized mail text is generated.

In this case, a grouping unit for grouping the plurality of mail recipients into one of a plurality of groups based on the personal information and the personalized information is provided, and the editing unit is the The personalized mail body may be generated for each group. Further, the analysis unit includes a topic analysis unit that analyzes a topic for each of a plurality of paragraphs forming the email body, and the editing unit has an interest in the analyzed topic and each email recipient's topic. Based on the above, the order of the paragraphs forming the mail body may be changed to include an order changing unit that determines the order of the paragraphs of the personalized mail text.

Further, the analysis unit includes a topic analysis unit that analyzes a topic for each of a plurality of paragraphs forming the mail body, and the editing unit responds to the analyzed topic and the topic of each mail recipient. A summarizing section may be provided for summarizing sentences forming paragraphs included in the mail body based on presence or absence of interest to generate paragraphs forming the personalized mail. Furthermore, the editing unit may include, as the editing process, an expression changing unit that changes the expression of at least a part of the mail text based on a preset condition. Further, the expression changing unit may change the ending of a sentence forming the mail body, with the gender of the mail recipient as the condition. Further, the expression changing unit may change the phrase of the sentence forming the mail body based on a word locally used in the area, with the area specified in advance as the condition. Furthermore, the expression changing unit may have a predetermined grade of the mail recipient as the condition,
The kanji notation of the kanji included in the mail body may be changed for each grade.

Further, in the control method of the mail generation support device for supporting the generation of the mail to be delivered to a plurality of mail recipients, an analysis process of analyzing the mail text of the mail manuscript using natural language processing, and each of the mails. Based on the personal information corresponding to the recipient, preset personalization information, and the result of the analysis, the email text is edited by using natural language processing so as to correspond to each email recipient, and the personalized email text is created. And an editing process for generating. In this case, a grouping process for grouping the plurality of mail recipients into one of a plurality of groups based on the personal information and the personalized information is provided, and the editing process is performed for each group. The personalized email body may be generated.

Further, the analysis step includes a topic analysis step of analyzing a topic for each of a plurality of paragraphs forming the mail text, and the editing step relates to the analyzed topic and the topic of each mail recipient. A step of changing the order of paragraphs forming the body of the mail to determine the order of paragraphs of the body of the personalized mail may be provided based on whether the person is interested or not. Further, the analysis step includes a topic analysis step of analyzing a topic for each of a plurality of paragraphs forming the email body, and the editing step determines whether there is interest in the analyzed topic and the topic of each mail recipient. A summarization process of summarizing sentences forming paragraphs included in the mail body to generate paragraphs forming the personalized mail may be provided.

Further, the editing process may include an expression changing process for changing the expression of at least a part of the mail text based on a preset condition as the editing process. In addition, the expression change process,
The ending of the sentence forming the mail body may be changed under the condition of the sex of the mail recipient.
Further, in the expression changing process, the words and phrases of the sentence forming the mail body may be changed based on a word locally used in the area, with the area specified in advance as the condition. Furthermore, the expression changing process may change the Kanji notation of the Kanji included in the mail body for each grade, with the grade of the mail recipient designated in advance as the condition.

Further, the mail delivery system includes a mail body of a mail manuscript, personal information corresponding to each mail recipient, a mail transmitting terminal for transmitting preset personalized information, the received personal information, and the personalized information. And an individual who creates a personalized email body by editing the email body using natural language processing so as to correspond to each email recipient based on the email body and the analysis result obtained by analyzing the email body. Personalization server and personalized mail corresponding to each of the mail recipients based on the personal information and the personalized mail text received from the personalization server and corresponding to the mail recipients via a network A mail server for delivering to the terminal is provided. According to the above configuration, the mail sending terminal sends the mail text of the mail original, personal information corresponding to each mail recipient, and preset personalization information to the personalization server. The personalization server edits the mail text corresponding to each mail recipient based on the received personal information, personalization information, the mail text, and the analysis result obtained by analyzing the mail text to personalize the mail text. Is generated and sent to the mail server. The mail server generates a personalized mail corresponding to each mail recipient based on the personal information and the body of the personalized mail received from the personalization server and delivers it to the terminal corresponding to the mail recipient via the network. To do.

Further, a control program for causing a computer to function as a mail generation support device for supporting the generation of a mail to be delivered to a plurality of mail recipients analyzes the mail text of the mail manuscript using natural language processing, Based on the personal information corresponding to each of the mail recipients, preset personalization information, and the result of the analysis, the mail text is edited corresponding to each of the mail recipients to generate a personalized mail text. , Is characterized.
In this case, based on the personal information and the personalized information, the plurality of mail recipients are grouped so as to belong to any one of a plurality of groups, and the personalized mail text is generated for each group. May be. Also, the topic is analyzed for each of a plurality of paragraphs forming the mail text, and the order of the paragraphs forming the mail text is changed based on the analyzed topic and the interest of each mail recipient in the topic. Alternatively, the order of the paragraphs of the personalized mail body may be determined.

Further, a topic is analyzed for each of a plurality of paragraphs constituting the mail text, and the paragraphs included in the mail text are constructed based on the analyzed topic and the interest of each of the mail recipients in the topic. The sentences may be summarized to generate the paragraphs that compose the personalized mail. Furthermore, as the editing process, the expression of at least a part of the mail text may be changed based on a preset condition. Further, the sex of the mail recipient may be set as the condition, and the ending of the sentence forming the mail body may be changed as the expression change. Further, a previously designated area may be used as the condition, and the phrase of the sentence constituting the mail body may be changed based on a word locally used in the area as the change of the expression. Furthermore,
The grade of the mail recipient designated in advance may be used as the condition, and the Kanji notation of the Kanji included in the mail body may be changed for each grade as the change of the expression. Further, each of the above control programs may be recorded in a recording medium.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Next, preferred embodiments of the present invention will be described with reference to the drawings. [1] Schematic Configuration of Mail Delivery System FIG. 1 shows a block diagram of a schematic configuration of the mail delivery system of the embodiment. The mail delivery system 10 is roughly divided into
A plurality of mobile terminals 11-1, 11-2 and a radio base station 12
1, a personal computer system 13, a general public line network 14, the Internet 15, a mail transmission terminal 16, a personalization server 17, and a mail server 18. Each of the mobile terminals 11-1 and 11-2 has a wireless communication function and functions as a mail receiving terminal for displaying a mail received via wireless on a display. Specifically, the mobile terminals 11-1 and 11-2 are mobile phone terminals,
PHS (registered trademark) terminal, PDA (Personal
Digital Assistant) and the like.

The radio base station 12 comprises mobile terminals 11-1 and 1-1.
It provides an interface function between 1-2 and the general public network 14. Personal computer system 1
Reference numeral 3 denotes a display (not shown) for displaying various information including mail, an input device (not shown) such as a keyboard for inputting various data, an output device (not shown) such as a printer for outputting various information, and a general public network. 14
Alternatively, a communication device (not shown) such as a modem or a terminal adapter for performing communication via the Internet 15 is provided. The general public line network 14 is composed of a general telephone line network or the like. Internet 15
Is composed of a plurality of computer networks. The mail sending terminal 16 is the mobile terminal 11-1, 11
-2 Or create a manuscript mail (mainly the manuscript mail text) for generating a distribution mail to be distributed to the personal computer system 13 and input personalization information for personalizing the manuscript mail and generating a distribution mail Then, the function of transmitting to the personalization server 17 is provided. The personalization server 17 provides a function of generating a distribution mail to be actually distributed to each individual based on the original mail and the personalization information transmitted by the mail transmission server 16 and transmitting it to the mail server 18.

The mail server 18 provides a function of distributing the distribution mail transmitted from the personalization server to the mobile terminals 11-1, 11-2 or the personal computer system 13 via the Internet 15 and the general public line network 14. To do. In this case, the mail sending terminal 16,
The personalization server 17 and the mail server 18 are actually
Microprocessor unit (MPU), R not shown
The computer includes an OM, a RAM, an external storage device, an input / output device, and a display device. Also,
The functions of the mail transmission terminal 16, the personalization server 17, and the mail server 18 are realized by a program executable by a microprocessor unit corresponding to each server.
In addition, such a program can be stored in a semiconductor memory, a CD-
It may be directly executed from a recording medium such as a ROM. Also,
It is also possible to install the program in an external storage device in advance and execute it.

[2] Schematic Configuration of Personalization Server FIG. 2 shows a schematic configuration block diagram of the personalization server 17.
The personalization server 17 is roughly classified into a grouping engine 21, a thesaurus database 22, a topic analysis engine 23, a personalization engine 24, a standardization database 25, a standardization engine 26, and a terminology database 2.
7 and a term selection engine 28. The grouping engine 21 receives a plurality of mails corresponding to the personal information based on the personal information transmitted from the mail transmission terminal 16, the personalization item information and the personalization method information, and the thesaurus read from the thesaurus database 22. Group people into multiple groups.
In this case, personal information includes gender, age group, address,
It includes multiple items such as work location, tastes (hobbies, objects of interest, etc.).

Further, the personalized item information is information for designating which of the items included in the personal information the grouping is based on. Further, the personalization method information is information for designating how to personalize the mail for the items designated by the personalization item law above, out of the plurality of items constituting the personal information. Specifically, the following personalization can be considered. Whether to summarize the email or not. If so, how much do you want to summarize? Whether to change the notation like the Kanji notation for each grade. How would you like to change the notation? Whether to paraphrase into another word. In other words, whether to use a simpler expression, even if it has a slightly different meaning.

The thesaurus database 22 stores a thesaurus that represents a semantic classification. This thesaurus is
Used by the grouping engine 21 when calculating the degree of similarity (corresponding to the closeness of the interest field) based on the words related to the interest field when grouping the mail recipients based on the interest field of each individual To be done. The thesaurus also includes a topic analysis engine 2 for classifying the content of each topic into a plurality of topics included in the manuscript mail.
Used by 3. It is also used to determine which meaning on the thesaurus the morphologically analyzed word belongs to.

FIG. 3 is an explanatory view of the conceptual structure of the thesaurus database 22. The thesaurus database 22
It has a tree structure representing the relationship between the upper level and the lower level of the meaning on the thesaurus, and each level is assigned a word or a table belonging to the meaning of the level. For example, FIG. 3 shows a case where the thesaurus database 22 is expressed by a tree structure of three layers, and the meanings on the highest thesaurus include “hobby” and “political relation”. The meanings on the middle thesaurus include "sports", "fashion", and "cars" as the lower thesaurus meanings of "hobby".
Also, "politics" is a subordinate thesaurus of "politics-related". In the meaning on the lower thesaurus, words belonging to the meaning on the upper thesaurus are stored as tables TB1 and TB2. Specifically, the words that belong to the meaning of the thesaurus below "sports" are "baseball", "soccer", "athletics",
... is stored as a table TB1. Further, as the words belonging to the meaning of the thesaurus lower than "fashion", "Japanese clothes / kimono", "jeans", ... Are stored as a table TB2.

The topic analysis engine 23 classifies a plurality of topics included in the manuscript mail into fields based on the thesaurus stored in the thesaurus database 22 as well as the appearance of the paragraph structure of the manuscript mail, and classifies the topics into topics. Performs processing to divide the original mail into multiple blocks. In this case, the topic analysis also includes morphological analysis and syntactic analysis. The personalization engine 24 controls the standardization engine 26 and the term selection engine 28 to personalize the mail body of the subject analyzed original mail based on the grouping information. The standardization engine 26 changes the notation of the original mail based on the standardization database 25, the personalization item information, and the personalization method information. In addition, when selecting terms, standardization is performed as a preprocessing to absorb fluctuations in the notation, if necessary. Term selection engine 28
On the basis of the term database 27, performs processing that cannot be performed only by document standardization in the standardization engine 26, for example, processing such as adding readings, adding descriptive text, and selecting words.

[3] Outline of Personalized Mail Delivery Process Next, an outline of the personalized mail delivery process in the mail delivery system 10 of the embodiment will be described with reference to the overall process flowchart of FIG. First, prior to personalization of a mail, the operator of the mail transmission terminal 16 which is the mail transmission source sets various information necessary for personalization (step S1). Then, the mail text of the mail (original mail) to be personalized is transmitted to the personalization server 17 (step S2). On the other hand, the personalization server 17 groups users who are mail recipients based on the set information (step S3). In parallel with this, personalization server 1
7 performs a morphological layer analysis of the email body (step S4) and a topic analysis of the email body (step S5), and executes personalization processing of the email body (step S6). Then, the personalization server 17 transmits the personalized mail text, personal information such as a mail address corresponding to each mail recipient, and grouping information to the mail server 18 (step S7). As a result, the mail server 18
Will send a mail to each mail recipient (step S8). Hereinafter, the processes of steps S1 to S8 will be described in detail.

[3.1] Settings for Personalizing Mail; Step S1 First, prior to personalizing the original mail, the operator of the mail transmitting terminal 16 which is the sender of the mail makes various settings for personalizing the mail (step S1). S1). In this case, personal information for transmitting the mail is required, but the personal information may be prepared individually by the user of the mail transmission terminal 16 or personal information prepared in advance by the operation manager of the personalization server. May be configured to be used. Examples of personal information to be prepared include name, email address, age (or date of birth), sex, hobby information, preference information, and the like. In this case, it is preferable to prepare a plurality of pieces of hobby information or taste information.
In addition, the same word as the thesaurus stored in the thesaurus database 22 is used. In addition, various personalization methods can be defined, but it is also possible to configure the settings in advance from a menu.

[3.1.1] Personalization Method Setting Menu Screen FIG. 5 shows an example of the personalization method setting menu screen. The personalization method setting menu screen of FIG. 5 is an example of setting the following five items as personalization methods. Kanji by grade: It is an item to instruct whether to control the kanji notation according to the grade. As selectable items,
For example, do not control Kanji notation, control Kanji notation for each grade "Control by grade", control Kanji notation based on either elementary school or lower or junior high school or higher There are three items of "control".
Practically, the words are paraphrased according to the Kanji notation selection database described later.

Word ending change by sex: This is an item for instructing whether or not the word ending of a sentence is changed according to gender. Generally, since there is not much difference in wording between men and women, when changing the ending, it is also an item to instruct whether or not to express further. As a selectable item, the ending is not changed according to gender.
And, "do" to change the ending by gender. Further, when "Yes" is selected to change the ending depending on the gender, "Ku" is allowed as a selectable item, which allows a loose expression. Practically, the words are paraphrased according to the sex paraphrasing database described later.

Change of phrase according to place of birth: This is an item for instructing whether to change the phrase of a sentence depending on the place of birth. That is, it is also an item for instructing whether to change the phrase using the dialect corresponding to the place of birth. As a selectable item, do not change the phrase depending on where you are from
And, "do" to change the phrase based on gender. Practically, the paraphrase is paraphrased according to a dialect paraphrase database described later. It should be noted that it is possible to perform the same word / phrase changing process not only in the place of origin but also in the case where the mail recipient desires or the mail sender desires.

Summary by preference: This is an item for instructing whether or not to summarize based on the preference information of each mail recipient. Do not summarize as a selectable item "No"
There is "do" to summarize. If you select "Yes" to further summarize, it is possible to set the summarization rate to how much (compressing the text) is summarized for topics that the relevant mail recipient is not interested in. ing.

Paragraph change by preference: This is an item for instructing whether to change the order of paragraphs based on the preference information of each mail recipient. The selectable items include “No”, which does not change the paragraph order, and “Yes”, which changes the paragraph order.

[3.2] Transmission of mail body data; step S2 Next, the user of the mail transmission terminal 16 transmits the mail body data corresponding to the mail body to the personalization server 17 via the Internet 15 (step S2). ). Practically, the user of the mail sending terminal 16 is
6 and the Internet 15, the mail body data is transmitted to the personalization server 17 together with the request for personalization of the mail. As a transmission method, FTP (File Transfer Protocol) is used as a protocol, and the mail body data and the like are transmitted in a file format, or SMTP.
It is possible to send in electronic mail format using (Simple Mail Transfer Protocol).

[3.3] User Grouping Process Next, the grouping engine 21 of the personalization server 17 performs a user grouping process (step S3). Less than,
The user grouping process will be described with reference to the user grouping process flowchart of FIG. First, the grouping engine 21 reads the personalization item, the personalization method, the personal information, and the grouping information received by the personalization server 17 from the mail transmission terminal 16 (step S1).
1). Next, the grouping engine 21 represents the elements included in the personalization item in the personal information in the vector space (step S12), and groups the elements in the vector space by the clustering technique (step S13).

Specifically, the grouping engine 21 is
In a multi-dimensional vector space with each item of personalized items as the axis,
Vectorize individual characteristics based on the personal information of each user who is the recipient of the mail. Next, using the existing clustering technology for automatic classification of elements in space, vectorized individual characteristics are combined into a group of multiple users with similar characteristics, and the users are grouped into multiple groups. Will be divided. In this case, the operator who manages the personalization server 17 can set the number of groups to be generated.

Next, the grouping engine 21 analyzes the generated personal information in each group to determine the characteristics of each group (step S14). Here, the determination of the group feature in the grouping engine 21 will be specifically described. In this case, regarding the characteristics of users belonging to the same group, a threshold value corresponding to personal information is determined in advance, and if the threshold value is exceeded, the characteristics are determined. Judge that you are doing. For example, as a threshold for determining a feature,
Set as follows. Age: When the threshold value is set to 80 [%], the age group to which 80 [%] or more individuals belong to the group is treated as the age group representing the characteristics of the group. Gender: When the threshold is set to 80 [%], the gender to which 80 [%] or more individuals belong to the group is treated as the gender representing the feature of the group. Preference: When the threshold is set to 50 [%], the preference to which 50 [%] or more individuals belong to the group is treated as the preference indicating the feature of the group.

Discrimination of the characteristics of the users belonging to the group A automatically created by clustering using the above threshold will be described. In this case, the total number of users belonging to group A is 1,000. Also,
The information for determining the characteristics based on the personal information of the users belonging to the group A is as follows. Age: under 20 years old = 820 people, under 30 years old = 100 people,
30 years old or older = 80 people Gender: 600 people, women = 400 people Preference: 600 people interested in baseball In this case, the characteristics of group A are "age characteristics are teens, tasteful. The feature is judged to be baseball. As for gender, neither male nor female exceeds the gender threshold = 80 [%], and therefore is not extracted as a group feature. The characteristics of the group to which the mail recipient belongs are determined by the above processing.

[3.4] Morphological analysis of mail text Next, morphological analysis of the mail text is performed to convert the text of the mail into sentences,
Divide into words (step S4). As the morphological analysis method in this case, various known methods can be applied. [3.5] Topic analysis of email body; step S5 Then, the personalization server 17 performs topic analysis of the email body (step S5). First, the outline processing of topic analysis will be described. FIG. 7 shows a general processing flowchart of topic analysis. Topic analysis engine 23 of personalization server 17
Uses the existing method of "semantic paragraph analysis, automatic discrimination" as disclosed in Japanese Patent Laid-Open No. 11-259521, for example, to block the sentence into blocks for each topic (step S21). Next, the topic analysis engine 23 determines an important word for each topic block (step S22). Then, the topic analysis engine 23 gives meaning to the words in the topic block based on the thesaurus (step S23). Further, the topic analysis engine 23 performs weighting according to the degree of importance of the word, and determines which field the topic block belongs to based on the word having the high degree of importance (step S24).

Here, a method of discriminating a topic will be described in detail. After being divided into a plurality of sentences by the automatic discrimination of the topic paragraph in step S21, the thesaurus database 22 is referred to for each sentence to check the meaning of the word on the thesaurus. In this case, the importance of words by the TFIDF method is used for weighting. FIG. 8 shows an example of a sentence to be a subject of discrimination. First, the words included in the sentence shown in FIG. 8, the frequency of appearance of each word, log (IDF), and the meaning (corresponding to topic) in the thesaurus of each word are obtained. An example is as follows. Word Frequency log (IDF) Meaning Sports 1 0 Sports Topic 10 (none) Major League 2 0.91 Baseball Ichiro 10 Baseball ………………………… Activity 2 0.91 (none) ……………… .. Here, the word frequency threshold value is 2 (times), so there are only two words to be calculated for topic detection: "major league" and "active". Next, the degree of importance of the topic corresponding to the word to be calculated for topic detection is calculated. In this case, the importance is given by the following equation. Importance = Word log (IDF) × Word appearance frequency In the above example, the calculation result is as follows. Word Meaning (Thesaurus) Importance Major League Baseball 1.82 Active (None) 1.82

By the way, in the above example, since the only word having a meaning is "major league" and the corresponding meaning is "baseball", the field to which the topic of the sentence shown in FIG. 8 belongs is "baseball". It is determined that there is. Similarly, when it is determined that a plurality of topics exist, the topic having the highest degree of importance is determined to be the topic of the sentence. Further, when there are a plurality of words having the same meaning, the sum of the degrees of importance of the plurality of words having the same meaning is set as the topic importance.
For example, a case where the calculation result of the degree of importance of a topic is as follows will be described. Word Meaning (on thesaurus) Importance Major league baseball 1.82 Ichiro baseball 1.53 In this case, the importance of the field "baseball" to which the topic belongs is 1.82 + 1.53 = 2.35.

The text of the mail often includes a greeting. Therefore, "greeting" is assigned as the meaning (on the thesaurus) to words that often appear in greetings (words that indicate the season, the time of day, phrases used in greetings, etc.). As for the meaning (on the thesaurus) of the words belonging to the sentence of which the topic is to be discriminated, if the significance of "greeting" is high on the meaning corresponding to other words (on the thesaurus), that is, the degree of greeting is If it is high, the sentence to be discriminated is judged to be a greeting. Then, the position of the greeting text existing at the beginning or the end of the mail body in the original mail is fixed, the greeting is excluded from the target for rearranging the order of the sentences to be described later, and the position is held.

[3.6] Personalization Processing of Email Body; Step S6 Next, the personalization processing of the email body will be described with reference to the flowchart of personalization processing of email body in FIG. In the personalization processing of the mail body (step S5), the personalization engine 24 receives the morpheme-analyzed and topic-divided mail body from the topic analysis engine 23 (step S31). Subsequently, the personalization engine 24 compares the topic corresponding to each topic block with the characteristics of the user group (interest in the topic) (step S32). That is, it is determined whether the topic corresponding to each topic block matches the topic of interest of each user group.

Next, the personalization engine 24 summarizes the sentence of the topic block as needed (step S33). In other words, for topics that are not of interest to each user group, even if they are described in detail, they cannot be read with interest, so a summary will be given. Here, the sentence summary process will be described in detail. In this case, the summarization is basically performed by the sentence unit, that is, the important sentence is extracted. When summarizing, first, in sentence units,
Which sentence is more important, that is, a sentence having a meaning closer to the meaning of the whole sentence before summarization is calculated by the TFIDF weighting method. Next, based on the result of this calculation, the sentences considered important are selected in order. Then, the ratio of the number of characters of the entire selected sentence to the number of characters of the entire sentence before the summarization is set to be closest to the desired summarization ratio, and the selected sentence is set as the summarized sentence.

More specifically, the importance of each word of the sentence is determined by using the TFIDF weighting method for the sentence whose morpheme has been analyzed. Although various calculation methods using TFIDF weighting have been proposed, in the following specific examples, a case will be described in which processing is performed under the following preconditions. The similarity between the whole sentence before summarization consisting of multiple sentences and each sentence is compared by the Euclidean distance. Words with an appearance frequency less than the word frequency threshold are not to be calculated, and this word frequency threshold is set to 2 (times). The summarization rate is 50 [%]. FIG. 10 shows an example of the entire sentence before being summarized. First, when summarizing, morphological analysis is performed on the entire sentence before the summarization, and the sentence is divided into sentences and words. FIG. 11 shows a division result obtained by dividing the entire pre-summarization sentence of FIG. 10 into sentence and word units. In the case of this specific example, it is divided into four sentences, that is, the first sentence to the fourth sentence.

Next, for each sentence of the first sentence to the fourth sentence, TFI
The DF weighting method is applied to calculate the similarity to the entire sentence. FIG. 12 shows the similarity to the whole sentence obtained by applying the TFIDF weighting method to each sentence after division.
As shown in FIG. 12, the similarity (score) of the first sentence is 0.
26, the similarity of the second sentence was 0.23, the similarity of the third sentence was 0.27, and the similarity of the fourth sentence was 0.08. In this case, the smaller the similarity value is, the closer the meaning is to the meaning of the whole sentence before summarization. Therefore, the meanings are arranged as follows. 4th sentence → 2nd
Sentence → 1st sentence → 3rd sentence

Here, the number of characters of the whole sentence before the summary and the number of characters of each sentence are as follows. Number of characters in the whole sentence before summarization: 141 (characters) Number of characters in the fourth sentence: 58 Number of characters in the second sentence: 28 Number of characters in the first sentence: 16 Number of characters in the third sentence: 39 Summarization rate of the next summary sentence that can be generated The result is as follows. Number of characters (characters) Summary rate (%) Select the 4th sentence: 58 41 Select the 4th sentence and the 2nd sentence: 86 60 Select the 4th sentence, the 2nd sentence and the 1st sentence: 102 72 Select the entire sentence: 141 100 From the above result, if the summarization rate exceeds the set summarization rate of 50 [%] and the closest case is selected as the summarization result, the sentences selected as the summarization result are the fourth sentence and the second sentence. Become.

However, at this point, the summarization rate is 50.
[%] Exceeded, so selected 4th sentence and 2nd sentence
Regarding the words in the sentence, which can be omitted, the term selection engine makes a determination using the omissible word / paraphrase word database included in the term database.

FIG. 13 shows an example of an abbreviation word / paraphrase word database. As shown in FIG. 13, optional words /
The paraphrasable word database 30 roughly includes paraphrase source data 31, paraphrase destination data 32, paraphrase destination part-of-speech data 33, and paraphrase level data 34. The paraphrase source data 31 is a word that can be included in the summary target sentence and stores a paraphrasable word. The paraphrase destination data 32 is the paraphrase source data 31.
The word of the paraphrase of the word corresponding to is stored. In addition, the part in which the word of the paraphrase is written as (NULL) indicates that the word is deleted at the time of paraphrase. Specifically, the words “always”, “almost”, and “pretty” are words that are deleted after paraphrasing. Also, the word "immediately" becomes "immediately" after being paraphrased.

When the word of the paraphrase is written as (NULL), the inflection ending, particle, etc. described corresponding to the word of the paraphrase are deleted. The paraphrase target part-of-speech data 33 deletes the inflection endings, particles, and the like described corresponding to the rewording source word when the reworded word is present, replaces it with the rephrased word, and then It is used to automatically change particles, etc. according to the word after paraphrase. The paraphrase level data 34 is data that stores a paraphrase level that indicates which word in which a plurality of eliminable words or paraphrasable words exist in the same sentence is preferentially omitted or paraphrased. That is, the whole sentence selected as the summary result (the fourth sentence and the second sentence in the above example) is deleted or paraphrased in descending order of paraphrase level (large numerical value in this embodiment), Paraphrasing is performed within a range in which the number of characters in the summary result sentence after deletion or paraphrasing does not become smaller than the number of characters corresponding to the summarization rate.

Further, the sentence after deletion or paraphrasing of each word is subjected to morphological analysis, connection check is performed based on the information of the word and part of speech, and if the result of the connection check is not valid, deletion or paraphrase is not performed. ing. As a result, the summary sentence obtained in the above example is as follows. “Yukata is something to wear in the hot season. The yukata fashion is changing these days.” The total number of characters in the summary sentences generated by the 4th and 2nd sentences is 83, and the summarization rate is 59 [ %]. Therefore, this summary rate is the set summary rate 50 [%].
Since the above requirement is satisfied, the obtained summary result sentence will be adopted as the summary sentence.

Next, the personalization server 17 selects terms used in the mail (title, mail body, etc.) (step S34). For example, in the setting process of step S1, when the control of ending change by sex is selected, the word is paraphrased according to the sex paraphrase database 40 shown in FIG. Specifically, when the gender-based paraphrase database 40 shown in FIG. 14 is used, if the auxiliary verb ending form “... is.” When the recipient is an elementary school boy or a junior high school boy, paraphrase it to "... In addition, when the mail recipient is an elementary school girl or a junior high school girl, it is paraphrased as "... is it?".

If a part of the mail such as the title and the text of the mail has the auxiliary verb ending form "... Let's do.", If the mail recipient is a boy in elementary school, "... In other words. In addition, when the mail recipient is a junior high school boy to a man of university or above (adult), it is paraphrased as “to yo”. If the mail recipient is an elementary school girl or a junior high school girl, it is paraphrased as "... In addition, in the setting process of step S1, in the case where it is selected that the phrase change according to the place of birth is controlled, that is, the phrase change using the dialect corresponding to the place of birth is selected, it is determined for each dialect shown in FIG. Paraphrases are paraphrased according to the dialect paraphrasing database 50 provided in.

Specifically, when the Hiroshima area dialect paraphrase database 50 shown in FIG. 15 is used, the word "I" (personal pronoun) is used in a part of the mail such as the title and mail text. If so, translate it into "I". If the word "thank you" (idiomatic phrase) is used, "Thank you"
In other words, Next, the personalization engine 24
The standardization engine 26 is controlled to refer to the standardization database to change the Kanji notation (step S35).

That is, in the setting process of step S1, the Kanji notation is controlled, that is, the Kanji notation is controlled for each grade, or the Kanji notation is set based on either the elementary school or lower or the junior high school or higher. When “Control only elementary school and below / junior high school and above” is selected, which is included in the standardized database 25 shown in FIG.
The standardization engine 26 paraphrases words based on the Kanji notation selection database 60 shown in FIG.
Specifically, if "Control by grade" is selected and the word "Activity" is used in a part of the email such as the title and email text, the first grader is the email recipient. In some cases, it is translated into the word "Katsuyaku". Similarly, in the second grade to the sixth grade in elementary school, the word "lively" is used. However, if you are a junior high school student or older, you will not be paraphrased and will be written as "active". Further, when "control only below elementary school / junior high school" is selected, control according to grade is not performed, and the same paraphrasing is performed if the mail recipient is in sixth grade or lower.

Next, the personalization server 17 determines whether or not the processing for all topic blocks has been completed (step S).
36). If it is determined in step S36 that the processing for all topic blocks has not been completed (step S36).
36; No), the process moves to step S32, the next topic block is acquired, and the same process is repeated. In the determination of step S36, when the processing for all the topic blocks (all paragraphs) is completed (step S36; Yes), the order of the topic blocks (paragraphs) is changed according to the preference of the user group to be processed (step). S37). Here, the process of changing the order of topic blocks (paragraphs) will be described with a specific example.

FIG. 17 shows an example of the mail text composed of four topic blocks (paragraphs) before the order change. As shown in FIG. 17, the mail body has a general greeting, the first paragraph, a fashion topic, a second paragraph, a sports topic, a third paragraph, and a general greeting message, the fourth paragraph. It consists of Here, the following two groups A and B are assumed as groups of mail recipients. Group A is a group that is interested in sports, and it is assumed that settings other than kanji learned in elementary school are set to change to kana notation or to perform processing such as wiggling. In this case, topics related to sports that are of interest should move toward the beginning, and topics that are not of interest (in the case of the above example, topics of fashion) should be 50.
[%] Shall be summarized.

FIG. 18 shows the mail text with the order changed and the notation changed corresponding to group A. As shown in FIG. 18, the third paragraph on sports is a new second paragraph. In addition, the second paragraph on the topic of fashion will be a new third paragraph, and
[%] Is summarized. Further, the word "great devil" is spelled out by the term selection engine 28 which refers to the term database, and the word "active" is indicated by the standardization engine 26 which refers to the standardization database according to the learning kanji of elementary school children. Has been replaced by Yaku. On the other hand, Group B has no topics of interest,
It is assumed that the notation is not changed. Also in this case, for topics that are not of interest (in the above example, topics of sports and fashion), 50
[%] Shall be summarized.

FIG. 19 shows the mail text that has been processed for group B. As shown in FIG. 19, the second paragraph on the topic of fashion and the third paragraph on the topic of sports
The paragraphs are both summarized by 50 [%]. Next, the personalization server 17 determines whether or not the processing has been completed for all user groups (step S38). In the determination of step S38, when it is determined that the processing has not been completed for all user groups (step S38; No), the process proceeds to step S32, and the same process is repeated thereafter.

[3.7] Process of Sending to Mail Server; Step S7 If it is determined in the determination of step S38 that the mail personalization process for all user groups is completed (step S38; Yes), the mail is sent. The mail body after personalization, personal information, and grouping information indicating which individual belongs to which group are transmitted to the server (step S7).

[3.8] Mail delivery; step S8 As a result, the mail server 18 receives the mail body corresponding to each group based on the received mail body after personalization completion, personal information and grouping information. A mail including address information and the like is generated, and mobile terminals 11-1 and 11-2, which are terminals of the mail recipient, personal computer 1
It will be distributed to 3

[4] Effects of the Embodiments In personalization processing, natural language processing such as standardization and term selection is performed, so that the text can be made more natural and finely personalized. Since the email is personalized after the topic analysis, it is possible to perform detailed personalization not only for the entire email text but also for each paragraph (topic). Since the mail recipients are grouped and only the mail body of each group, personal information and grouping information indicating how the grouping is performed are sent from the personalization server to the mail server, the personalization server -The amount of communication between mail servers can be reduced. In addition, since personalization of mail is not performed for each person by this grouping, effective personalization of mail can be performed while reducing the load on the personalization server.

Further, since the personalization of the mail is performed on the personalization server side and the mail transmission terminal does not need to have the personalization function of the mail, the user of the mail transmission terminal may be concerned about capital investment, maintenance and the like. You can easily take advantage of email personalization without having to. On the other hand, the recipient of the email does not need to perform any processing for personalizing the email, and can receive the email corresponding to the taste of the individual based on the personal information already registered. It becomes possible to obtain the information to do.

[5] Modification of Embodiment [5.1] First Modification In the above description, the number of groups to be personalized for mail could not be extended in detail. If the number of conditions is further increased and more groups are formed, it is possible to personalize the mail more suitable for individuals belonging to each group.

[5.2] Second Modification In the above description, mails are personalized for each group, but personalized information that does not depend on groups for mails that are further personalized (Name, birthday, etc.) is sent to the personalization server or mail server in advance, and a message is added to each mail in the personalization mail server, or the mail title is changed in the mail server, and a subtitle is added. Alternatively, it can be configured to attach a predetermined attachment file. For example, it is possible to add "Notice to Mr. XX who is celebrating his birthday this month" as a subtitle.

[5.3] Third Modification In the above description, personalization of mail was performed for each group. However, if there is only one group member, personalization of mail is performed for each individual. This is also possible by the same method.

[5.4] Fourth Modification In the above description, the mail server, the personalization server, and the mail transmission terminal are provided independently of each other, but the mail server and the mail transmission terminal are the same company. It is also possible to configure so as to belong to.

[5.5] Fifth Modification In the above description, the case where the ending or the phrase is changed based on the sex, the grade, and the region where the student is born is described. However, various processes can be performed if various paraphrase databases are provided in advance. That is, as an editing process, at least a part of the expression of the mail body is changed based on a preset condition (replacement of a phrase, change of notation of a phrase, change of ending, and combination thereof). It is also possible to configure (expression change process). In this case, various settings are possible, such as changing the expression at random with a certain probability, prohibiting the expression change depending on the desired setting of the mail recipient. Examples of expression changes include katakana language notation, changing fonts and colors, paraphrasing technical terms depending on the field of expertise of the mail recipient, and substituting specific words within a range that can be understood. It is possible to turn the character into a flat character.

[0064]

According to the present invention, while reducing the burden and equipment load on the mail sender side, the mail text is edited and personalized using natural language processing so as to correspond to the mail recipient, and the mail is delivered. Is possible. In this case, since natural language processing is used, more natural and detailed personalization can be performed, so that a meaningful email can be generated and delivered to both the email sender and the email recipient. Further, topic analysis can be performed to make detailed personalization for each topic. Furthermore, since the personalization server and the mail server can be separately configured, maintenance of each server becomes easy.

[Brief description of drawings]

FIG. 1 is a schematic configuration block diagram of a mail delivery system according to an embodiment.

FIG. 2 is a schematic configuration block diagram of a personalization server.

FIG. 3 is an explanatory diagram of a conceptual structure of a thesaurus database.

FIG. 4 is an overall processing flowchart of personalized mail delivery processing.

FIG. 5 is a diagram showing an example of a menu screen for personalization method setting.

FIG. 6 is a user grouping process flowchart.

FIG. 7 is a schematic processing flowchart of topic analysis.

FIG. 8 is a diagram illustrating an example of a sentence that is a target of a topic determination.

FIG. 9 is a flowchart of a personalization process of a mail body.

FIG. 10 is a diagram showing an example of an entire sentence before being summarized.

FIG. 11 is a diagram for explaining a division result obtained by dividing the entire sentence before the summarization in FIG. 10 into sentence and word units.

FIG. 12 is a diagram for explaining the similarity to the entire sentence obtained by applying the TFIDF weighting method to each sentence after division.

FIG. 13 is a diagram for explaining an example of an abbreviation word / paraphrase word database.

FIG. 14 is an explanatory diagram of a sex paraphrase database.

FIG. 15 is an explanatory diagram of a dialect paraphrase database.

FIG. 16 is an explanatory diagram of a Chinese character notation selection database.

FIG. 17: Four topic blocks (paragraphs) before the order change
It is a figure explaining an example of the mail text comprised by.

FIG. 18 is an explanatory diagram of a mail body obtained by changing the order and the notation of the mail body shown in FIG. 17 in correspondence with group A.

FIG. 19 is an explanatory diagram of a mail body obtained by changing the order of the mail body shown in FIG. 17 corresponding to group B.

[Explanation of symbols]

10 …… Mail delivery system 11-1, 11-2 ... Mobile terminal 13 ... Personal computer system 14 ... General public network 15 ... Internet 16 …… Email sending terminal 17 ... Personalization server 18: Mail server 21 ... Grouping engine 22 ... Thesaurus database 23 ... Topic analysis engine (analysis part) 24 ... Personalization engine (editing department) 25 ... Standardized database 26: Standardization engine (editing department) 27 ... Term database 28 ... Term selection engine (Editor) 40 ... Database for paraphrasing by gender 50 ... Hiroshima dialect paraphrase database 60 …… Kanji notation selection database

Claims (26)

[Claims] 1. A mail generation support device for supporting generation of a mail to be delivered to a plurality of mail recipients, an analysis unit for analyzing a mail text of a mail manuscript using natural language processing, and each of the mail recipients. Based on the corresponding personal information, preset personalization information, and the result of the analysis, the email body is edited using natural language processing for each email recipient to generate a personalized email body. An e-mail generation support device, comprising: 2. The mail generation support device according to claim 1, wherein a grouping unit that groups the plurality of mail recipients into one of a plurality of groups based on the personal information and the personalized information. The mail generation support device, wherein the editing unit generates the personalized mail body for each group. 3. The mail generation support device according to claim 1, wherein the analysis unit detects a turning point of a topic in the entire email body and analyzes a topic included in the email body. A topic analysis unit is provided, and the editing unit changes the order of the paragraphs forming the email body based on the analyzed topic and presence / absence of interest in the topic of each of the email recipients, and paragraphs of the personalized email body. An email generation support device, comprising: an order changing unit that determines the order of. 4. The mail generation support device according to claim 1, wherein the analysis unit detects a turning point of a topic in the entire email body and analyzes a topic included in the email body. A topic analysis unit, wherein the editing unit summarizes sentences forming paragraphs included in the email body based on the analyzed topic and presence or absence of interest in the topic of each of the email recipients to personalize the email. An email generation support device, comprising: a summarization unit that generates paragraphs that compose the. 5. The mail generation support device according to claim 1, wherein the editing unit changes at least a part of the expression of the mail text based on a preset condition as the editing process. An email generation support device, comprising: 6. The mail generation support device according to claim 5, wherein the expression changing unit changes the ending of a sentence forming the mail body under the condition that the gender of the mail recipient is the condition. Mail generation support device. 7. The mail generation support device according to claim 5, wherein the expression changing unit sets the mail text based on a word that is locally used in the area with a predetermined area as the condition. A mail generation support device characterized by changing words of a constituent sentence. 8. The e-mail generation support device according to claim 5, wherein the expression changing unit uses the kanji of the kanji included in the mail body for each of the grades, with the grade of the mail recipient specified in advance as the condition. A mail generation support device characterized by changing notation. 9. A method of controlling a mail generation support device for supporting the generation of a mail to be delivered to a plurality of mail recipients, the analysis process of analyzing a mail text of a mail manuscript using natural language processing, and each of the mails. Based on the personal information corresponding to the recipient, preset personalization information, and the result of the analysis, the email text is edited by using natural language processing so as to correspond to each email recipient, and the personalized email text is created. A method of controlling a mail generation support device, comprising: an editing process for generating a mail. 10. The control method for a mail generation support device according to claim 9, wherein a plurality of the mail recipients are grouped into one of a plurality of groups based on the personal information and the personalized information. A method of controlling a mail generation support device, comprising: a grouping process, wherein the editing process generates the personalized mail body for each group. 11. The control method for a mail generation support device according to claim 9 or 10, wherein the analysis step detects a turning point of a topic in the entire body of the mail, and a topic included in the body of the mail. And a step of analyzing the topic, wherein the editing step changes the order of the paragraphs forming the mail text based on the analyzed topic and the interest of each of the mail recipients in the personalized mail. A method of controlling a mail generation support device, comprising: an order changing process that determines the order of paragraphs in the text. 12. The control method of the mail generation support device according to claim 9, wherein the analysis step detects a turning point of a topic in the entire email body, and a topic included in the email body. A topic analysis step of analyzing the above, wherein the editing step summarizes sentences constituting the paragraphs included in the email body based on whether the analyzed topic and each email recipient have an interest in the topic, and A control method for a mail generation support device, comprising: a summarization process for generating paragraphs constituting a personalized mail. 13. The control method for a mail generation support device according to claim 9 or 10, wherein in the editing process, at least a part of the mail text is edited based on a preset condition as the editing process. A method of controlling a mail generation support device, comprising: an expression changing process for changing an expression. 14. The method of controlling a mail generation support device according to claim 13, wherein the expression changing step changes the ending of a sentence forming the mail body, with the gender of the mail recipient as the condition. A control method of a characteristic mail generation support device. 15. The method of controlling a mail generation support device according to claim 13, wherein the expression changing process is based on a word locally used in the area, with a predetermined area as the condition. A method of controlling a mail generation support device, characterized in that a phrase of a text constituting a mail body is changed. 16. The method of controlling a mail generation support device according to claim 13, wherein the expression changing process is included in the mail text for each of the grades, with the grade of the mail recipient specified in advance as the condition. A method for controlling a mail generation support device, characterized by changing the kanji representation of kanji. 17. A mail text of a mail manuscript, personal information corresponding to each mail recipient, a mail sending terminal for sending preset personalization information, the received personal information, the personalization information and the mail text, and A personalization server that edits the email body using natural language processing in correspondence with each of the email recipients based on an analysis result obtained by analyzing the email body to generate and send a personalized email body. And a personalized mail corresponding to each of the mail recipients based on the personal information and the body of the personalized mail received from the personalization server, and a terminal corresponding to the mail recipient via a network. A mail delivery system characterized by having a mail server for delivery to the user. 18. A control program for causing a computer to function as a mail generation support device for supporting the generation of a mail to be delivered to a plurality of mail recipients, wherein the mail text of a mail manuscript is analyzed using natural language processing, Based on the personal information corresponding to each of the mail recipients, preset personalization information, and the result of the analysis, the mail text is edited corresponding to each of the mail recipients to generate a personalized mail text. , A control program characterized by the following. 19. The control program according to claim 18, wherein a plurality of the mail recipients are grouped into one of a plurality of groups based on the personal information and the personalized information, and each of the groups is grouped. A control program for generating the personalized mail body. 20. The control program according to claim 18 or 19, wherein a turning point of a topic in the entire email body is detected, a topic included in the email body is analyzed, and the analyzed topic and A control program, characterized in that the order of the paragraphs of the personalized mail body is determined by changing the order of the paragraphs forming the mail body based on whether each of the mail recipients is interested in the topic. 21. The control program according to claim 18 or 19, wherein a turning point of a topic in the entire email body is detected, a topic included in the email body is analyzed, and the analyzed topic and A control program, characterized in that paragraphs included in the mail body are summarized to generate paragraphs forming the personalized mail based on whether each mail recipient has an interest in the topic. 22. The control program according to claim 18 or 19, wherein, as the editing processing, at least a part of the expression of the mail body is changed based on a preset condition. Control program. 23. The control program according to claim 22, wherein the sex of the mail recipient is used as the condition, and the ending of a sentence forming the mail body is changed as the expression is changed. . 24. The control program according to claim 22, wherein a text that composes the mail text is based on a word that is locally used in the area as a change of the expression, with the area designated in advance as the condition. A control program characterized by changing the phrase. 25. The control program according to claim 22, wherein the grade of the mail recipient designated in advance is used as the condition, and the Kanji notation of the Kanji included in the mail text is changed for each grade as the expression is changed. A control program characterized by: 26. A recording medium on which the control program according to any one of claims 18 to 25 is recorded.