JP4635454B2 - Evaluation apparatus and method - Google Patents

Description

  The present invention relates to an evaluation apparatus and method for evaluating the influence of a certain organization on other organizations based on information transmitted between a plurality of organizations.

Currently, as an organizational communication (organizational communication) method using information technology (IT), e-mail, a mailing list using an e-mail system, and an electronic bulletin board using the WWW (World Wide Web) are generally used. Has been used.
Since the messages transmitted by these organization communication means are stored as message logs, the stored messages can be analyzed and the analysis results can be used as necessary.

For example, Patent Documents 1 to 7 disclose evaluation methods for companies, organizations, brands, and the like.
However, none of the methods disclosed in Patent Documents 1 to 7 can evaluate how a concept propagates between a plurality of organizations using the accumulated message log.
Japanese Patent Laid-Open No. 2001-3389 JP 2001-350911 A JP 2002-207844 A JP 2003-141269 A JP 2003-157279 A JP 2003-187052 A JP 2003-216563 A

  The present invention has been made from the background described above, and is an evaluation apparatus that can objectively evaluate how a concept propagates between multiple organizations and what kind of influence it has. And an object thereof.

  For the above purpose, the evaluation apparatus according to the present invention is transmitted between a survey means for investigating the mode of influence exerted between a plurality of evaluation target units that can be evaluated and the plurality of evaluation target units. , Analyzing the transmission information data each including one or more concepts and determining temporal transitions in the plurality of evaluation target units of the concepts included in the transmission information data. Correction for correcting the evaluation of the influence of the evaluation target unit on the other evaluation target unit based on the evaluation means for evaluating the influence on the other evaluation target unit and the mode of the influence obtained as a result of the investigation Means.

  Preferably, the investigating means investigates, as an aspect of the influence, which of the other evaluation object units has affected each component of the evaluation object unit, and the evaluation means exerts an influence. The influence from the first evaluation target unit to the affected second evaluation target unit is evaluated, and the correction means determines the second evaluation target unit from the first evaluation target unit based on the result of the investigation. Correct the assessment of the impact on the unit being assessed.

  Preferably, the correcting means is based on the result of the survey between the first evaluation target unit and the second evaluation target unit that is directly influenced by the first evaluation target unit. Based on the first degree of influence received from the first evaluation target unit and the transmission information data, the overlapping constituent element is between the first evaluation target unit and the second evaluation target unit. The overlapping component obtains the second degree of influence exerted on the second evaluation target unit, and the duplicate component constitutes the obtained first influence and the obtained first degree. Based on the degree of influence of 2, the evaluation of the influence on the second evaluation target unit is corrected from the first evaluation target unit.

  Preferably, the correction unit calculates a coefficient obtained by multiplying the obtained first influence degree and the obtained second influence degree from the first evaluation target unit. The evaluation of the influence on the second evaluation target unit is corrected from the first evaluation target unit by multiplying the numerical value indicating the evaluation of the influence on the second evaluation target unit.

  Preferably, the correction means determines the number of the evaluation target units interposed in the propagation path of the influence from the first evaluation target unit to the second evaluation target unit based on the result of the investigation. And, based on the result of the investigation, an evaluation target unit intervening in the propagation path of the influence, and an average value of the degree of influence between the evaluation target units directly affecting the Based on the number of the evaluation units interposed in the propagation path of the influence and the average value of the obtained degree of influence, an evaluation of the influence from the first evaluation target unit to the second evaluation target unit is performed. to correct.

  Preferably, the correction means uses a coefficient obtained as a power of a numerical value corresponding to the number of the evaluation units interposed in the propagation path of the obtained influence of the average value of the obtained influence degree. The evaluation of the influence on the second evaluation target unit is corrected from the first evaluation target unit by multiplying the numerical value indicating the evaluation of the influence on the second evaluation target unit from the evaluation target unit. To do.

  Further, the evaluation method according to the present invention investigates the mode of influence exerted between a plurality of evaluation target units that can be evaluated, and is transmitted between the plurality of evaluation target units, each of which includes one or more concepts. Is analyzed, and the transition of temporal influence in the plurality of evaluation target units of the concept included in the transmission information data is determined, so that the evaluation target unit becomes another evaluation target unit. The influence exerted is evaluated, and the evaluation of the influence of the evaluation target unit on the other evaluation target units is corrected based on the aspect of the influence obtained as a result of the investigation.

  Further, the program according to the present invention is transmitted between the investigation step for investigating the aspect of the influence exerted between a plurality of evaluation target units that can be evaluated, and the plurality of evaluation target units, each of which is one or more. By analyzing the transmission information data including the concept of and determining the transition of the temporal influence of the concept included in the transmission information data in the plurality of evaluation target units, the evaluation target unit becomes the other evaluation target. An evaluation step for evaluating the influence on the unit, and a correction step for correcting the evaluation of the influence of the evaluation target unit on the other evaluation target unit based on the aspect of the influence obtained as a result of the investigation. Let it run.

  According to the present invention, it is possible to analyze a message transmitted within an organization and between organizations and evaluate how a concept propagates between a plurality of organizations over time.

[Background of the present invention]
In order to facilitate understanding of the present invention, prior to the description of the embodiment, first, the background that led to the present invention will be described.
Until now, the value of an organization (whether it is an organization with an entity such as a company department or a virtual organization such as a mailing list, as long as it is an evaluation target unit) It has been evaluated by official transactions and services provided outside the company or sales to the outside.
For example, in a company, organizations that are functionally differentiated with respect to problems to be solved arranged hierarchically are arranged hierarchically, and the problem is solved by transmitting a command to such an organization.

Within such a corporate structure, the assessment of an organization's value is, for example, the input of resources such as people, goods and money that are purely invested in the organization, and the resulting economic value and services. It has been evaluated as the difference and ratio with the output.
However, it has been pointed out from both the academic and actual business aspects that the value of an organization cannot be assessed only from the perspective of the difference and ratio between input and output.

Also, in the field of business administration, it is pointed out that such a problem solving method using a hierarchical organization cannot quickly respond to various requests of customers.
In response to such problems, IT (Information Technology) has been introduced, the organization within the company has a flat structure, autonomous activities are recommended to each organization, and resources are allocated to solve the problem. A solution to flexibly change and optimize is proposed.

However, when the structure of the organization is made flat in this way, it becomes extremely difficult to evaluate the value of the organization and its members based on the difference and ratio between the input and the output.
Because, as described above, in a flat organization, the structure changes flexibly, so there is no organization that aims to solve any problems from the beginning, and the organization composition of the organization always changes. Or one member may formally or informally belong to more than one organization, and after the problem is solved, the organization itself will be dissolved. This is because it is difficult to evaluate what kind of input was made and what kind of output it was.

The present invention focuses on communication within an organization and between organizations (organizational communication) regardless of whether it is formal or informal in order to evaluate the value of the organization.
Specifically, the present invention pays attention to what range and scale of communication that occurs (transmitted) in one organization is used in other organizations. Evaluate the value of
In other words, the present invention pays attention to organization communication and determines the value of a certain organization using the influence of the organization on another organization.

Focusing on these points, the value of organizations and organizational communications that did not produce direct economic value, such as providing information and services not officially imposed on a certain organization, or that were difficult to evaluate. Can alternatively be evaluated.
More specifically, when an organization is solving a problem, by analyzing the content of the communication in this organization, information in one organization can be communicated to other organizations and their members, Focusing on what is used for the activities of other organizations and their members, it is possible to objectively evaluate what is different between the value evaluation of the organization and the value evaluation of the organization communication.
For example, in a company, not only the value of the department that is actually making profits, but at first glance it seems that the value is low because it does not make profits, but in fact it is beneficial to many departments of the company and is indirectly Since the value of departments that are linked to large profits can be quantitatively evaluated, understanding the value of the organization from this perspective is useful for appropriate allocation of management resources such as budgets and personnel. Can contribute to corporate performance.

The applicant of the present application has already applied for an invention that supports various analyzes of organization communication performed on a daily basis as Japanese Patent Application No. 2001-275808 (organization communication analysis apparatus and method; Application 1).
The invention according to application 1 is a factor that activates or inactivates organizational communication by performing quantitative analysis on organizational communication such as a mailing list / electronic bulletin board and associating the analysis result with qualitative information. In an exploratory analysis, the organization communication operation is realized based on the analysis results.

The present invention is a further development of the invention according to the application 1 from the viewpoints described so far, and realizes an objective value evaluation of an organization by performing an analysis on the organization communication.
Furthermore, the present invention evaluates and realizes how a certain concept propagates between organizations (groups) and what kind of influence is given to each organization (group).

Examples of organizational communication means (media) include verbal, telephone, video telephone systems, and computer networks (e-mail, electronic bulletin board, chat, instant message, etc.).
In order to realize the present invention, it is premised that the communication performed through these media is recorded.
In the following, for the sake of concreteness and simplification of explanation, a case where an electronic mail is used in a company is taken as a specific example.
When e-mail is used for communication, generally the message in the natural language included in the e-mail, the sender and receiver of the e-mail, the date and time when the e-mail was transmitted, etc. Can be logged.

Distribution of electronic mail on the Internet is performed via a transmission server called an SMTP (Simple Mail Transfer Protocol) server.
Therefore, in the company, by recording the message log of the e-mail with the SMTP server in a unified manner, when and who communicated with whom and what kind of content for all employees in the company Can be grasped.
The present invention is configured to evaluate the value of an organization in a company by analyzing a message log of an SMTP server that distributes electronic mail in the company, for example.
Furthermore, the present invention is configured to analyze the message log described above and evaluate how a concept propagates and affects between organizations in a company.

[Embodiment]
Embodiments of the present invention will be described below.

[Network system 1]
FIG. 1 is a diagram illustrating a configuration of a network system 1 to which an evaluation method according to the present invention is applied.
In the following drawings, substantially the same constituent parts are denoted by the same reference numerals.
The network system 1 is, for example, a wide area network (WAN) constructed across a plurality of offices in the same company, and, as shown in FIG. 1, a plurality of organizations (groups) 1 to 1 that can be evaluated. The n departmental systems 2-1 to 2-n (n ≧ 2) and the analyzer 3 are connected via the network 100.
Hereinafter, in the case where any one of a plurality of components such as the departmental systems 2-1 to 2-n is indicated without being specified, it is simply abbreviated as the departmental system 2.
In each departmental system 2, for example, client computers 20-1 to 20 -m (m ≧ 1) used by m members of each organization (group) and the server 24 are connected via the departmental LAN 102. The configuration is taken.

[Hardware configuration]
FIG. 2 is a diagram illustrating a hardware configuration of the client computer 20, the server 24, and the analysis / evaluation apparatus 3 illustrated in FIG.
As shown in FIG. 2, the client computer 20, the server 24, and the analysis device 3 include a main body 200 including a CPU 202 and a memory 204, a display / input device 206 including a liquid crystal display, a keyboard and a mouse (not shown), an HDD It comprises a recording device 208 such as a CD device, and a communication device 212 that communicates with the network 100 and the departmental LAN 102.
That is, the client computer 20, the server 24, and the analysis device 3 include components as a general computer that can communicate via a network.

[Client program 22]
FIG. 3 is a diagram showing the configuration of the client program 22 that runs on the client computer 20 shown in FIGS.
As shown in FIG. 3, the client program 22 includes a user interface unit (UI unit) 220, a mail program 222, a web browser 224, and a LAN communication control unit 226.
For example, the client program 22 is supplied to the recording device 208 of the client computer 20 via the recording medium 210, loaded into the memory 204, and executed.
The client program 22 provides a mail transmission / reception function and a WWW browsing function to members (users) of an organization (group) that uses the client computer 20 by using these components.

In the client program 22, the UI unit 220 receives a user operation on the display / input device 206 (FIG. 2) and controls processing of each component of the client program 22.
In addition, the UI unit 220 displays the electronic mail received by the mail program 222 and the data from the WWW received by the web browser 224 to the user.

The mail program 222 provides the user of the client computer 20 with a function for sending and receiving electronic mail.
The web browser 224 provides a WWW browsing function to the user of the client computer 20.
The LAN communication control unit 226 is a communication node that collectively refers to other client computers 20 or servers 24 (communication main components in the same organization (group) via the departmental LAN 102 (FIG. 1) and the network 100. And communication with communication nodes of other organizations (groups).

[Server program 26]
FIG. 4 is a diagram showing the configuration of the server program 26 that operates on the server 24 shown in FIGS.
As shown in FIG. 4, the server program 26 includes a LAN communication control 260, a network communication control 262, a mail server program 264, a web server 266, a log management unit 268, and a log database (log DB) 270.
As with the client program 22 (FIG. 3), the server program 26 is supplied to the recording device 208 (FIG. 2) of the server 24 via the recording medium 210 (FIG. 1), loaded into the memory 204, and executed.

  With these components, the server program 26 provides a mail server function to the client computers 20 (members) belonging to the same departmental system 2 (organization (group)), and is the same or different departmental system 2. A WWW server function is provided to the client computer 20 (member) of (organization (group)).

In the server program 26, the LAN communication control 260 performs communication control with the departmental LAN 102 (FIG. 1).
The network communication control 262 performs communication control with the network 100.
The mail server program 264 implements a mail server function.
The web server 266 implements a WWW server function.

FIG. 5 is a diagram illustrating an e-mail message log recorded in the log DB 270 illustrated in FIG. 4.
The log management unit 268 records communication (communication) performed by the mail server program 264 and the web server 266 as a message log in the log DB 270 according to control data input from the analysis / evaluation apparatus 3 via the network communication control 262. To do.
Further, the log management unit 268 transmits the message log recorded in the log DB 270 to the analysis / evaluation apparatus 3 via the network communication control 262 and the network 100 according to the control of the analysis / evaluation apparatus 3.
The message log recorded by the log management unit 268 includes an ID (communication ID) attached to each e-mail, identification data (ID) of the side that sent and received the e-mail, and e-mail message. Contents (communication contents), the date and time when the electronic mail is transmitted, and the like are recorded.

[Analysis / Evaluation Program 30]
FIG. 6 is a diagram showing a configuration of a first analysis / evaluation program 30 that operates on the analysis / evaluation apparatus 3 shown in FIGS.
As shown in FIG. 6, the first analysis / evaluation program 30 includes a network communication control 300, a log collection / management unit 302, a log DB 304, an analysis / evaluation unit 4, an analysis / evaluation result DB 308, an organization / personal DB 310, and a UI. Part 312.
Similar to the client program 22 (FIG. 3) and the server program 26 (FIG. 4), the analysis / evaluation program 30 is supplied to the recording device 208 of the analysis / evaluation device 3 via the recording medium 210 and loaded into the memory 204. And executed (the same applies to the following programs).

With these components, the analysis / evaluation program 30 collects e-mail message logs, WWW browsing logs, and the like from the respective servers 24 (FIG. 1) of the departmental systems 2-1 to 2-n. Analyzing message logs, etc., and evaluating the value of the organization (group).
It should be noted that even if the WWW log is analyzed, the value evaluation of the organization (group) according to the present invention can be realized. However, as described above, in the description of the embodiment, the analysis of the email message log is performed. A specific example is a method of evaluating the value of an organization (group).

In the analysis / evaluation program 30, the network communication control 300 performs communication control with the network 100.
The UI unit 312 accepts a user operation on the display / input device 206 and controls processing of each component of the analysis / evaluation program 30.
Further, the UI unit 312 displays the log stored in the log DB 304 and the analysis result / evaluation result stored in the analysis / evaluation result DB 308 on the display / input device 206 in accordance with a user operation.

FIG. 7 is a diagram showing organization information stored in the organization / personal DB 310 shown in FIG.
FIG. 8 is a diagram showing personal information stored in the organization / personal DB 310 shown in FIG.
The organization / individual DB 310 includes organization information (FIG. 7) of organizations (groups) 1 to n (FIG. 1) and systems (groups) 1 to ( The personal information (FIG. 8) of the members 1 to m of group n is stored.

As shown in FIG. 7, the organization / personal DB 310 includes, as organization information of organizations (groups) 1 to n, identifiers (organization IDs) of organizations (groups) 1 to n, organization names, organization forms, and organizations (groups). When there is a superior organization (group) in the period (existence period) in which the entity exists, and the organizations (groups) 1 to n, the superior organization (group) is stored.
As shown in FIG. 8, the organization / individual DB 310 includes, as personal information of members 1 to m of organizations (groups) 1 to n, identifiers (personal ID / employee ID), names, and emails of the members 1 to m. The address and the organization ID (FIG. 7) of the organization (group) to which the members 1 to m belong are stored.
In the organizational form shown in FIG. 7, the official organization (group) as described in the organization chart of the company, and a plurality of official organizations (groups) are gathered to achieve a certain purpose. Project organization (group), organization (group) such as community based on voluntary participation, organization (group) such as information sharing group with the same interests (normal organization, project, community, etc.) Stored.

FIG. 9 is a diagram showing organization communication information (organization communication information) stored in the log DB 304 by the log collection / management unit 302.
Note that FIG. 9 shows a case where there is one sender and one receiver, but by changing the data structure shown in FIG. 9 appropriately, a one-to-many email by a mailing list is shown. The organization communication information can also be stored in the log DB 304.
The log collection / management unit 302 collects the message logs shown in FIG. 5 from each of the servers 24 of the departmental systems 2-1 to 2-n via the network 100 and stores them in the log DB 304.

That is, the log collection / management unit 302 refers to the organization / personal DB 310, associates the message log obtained from the server 24 with the organization information shown in FIG. 7 and the personal information shown in FIG. Thus, the organization communication information (FIG. 9) is created and stored in the log DB 304.
In other words, the log collection / management unit 302 searches the personal information (FIG. 8) stored in the organization / personal DB 310 using the e-mail sender's e-mail address and the receiver's e-mail address, and sends the e-mail. i's personal ID (employee ID) and personal ID (employee ID) of member j (i, j = 1 to m) who received the e-mail, and personal ID of the member i who sent the e-mail (employee) ID) is the sender ID, and the individual ID (employee ID) of the member j who has received the e-mail is the receiver ID.

Further, the log collection / management unit 302 searches the organization information (FIG. 7) stored in the organization / personal DB 310 using the sender ID and the sender ID, and the organization to which the member i who sent the email belongs belongs. The organization ID of (group) p is set as the sending organization ID, and the organization ID of the organization (group) q (p, q = 1 to n) to which the member j who has received the e-mail belongs is set as the receiving organization ID.
Further, the log collection / management unit 302 adds an identifier (organization communication ID) to each piece of information obtained as a result of the above search, adds the message content of the e-mail as communication content, The sent date / time and organization form are added, and the organization communication information shown in FIG. 9 is created and stored in the log DB 304.

[Analysis / Evaluation Department 4]
The analysis / evaluation unit 4 shown in FIG. 6 will be described below.
FIG. 10 is a diagram illustrating a configuration of the first analysis / evaluation unit 4 illustrated in FIG. 6.
As shown in FIG. 10, the analysis / evaluation unit 4 includes an evaluation target / range specification unit 400, a member classification unit 402, a preprocessing unit 420, a totaling unit 422, a feature extraction unit 424, a section determination unit 426, and an influence degree calculation unit. 428 and a result management unit 440.

The analysis / evaluation unit 4 analyzes the organization communication information (FIG. 9) stored in the log DB 304 using these components, and a certain concept propagates between organizations (groups) by any route. Evaluate the degree of impact on each (group).
The analysis / evaluation unit 4 evaluates the value of each organization by evaluating the propagation path of such a concept and the degree of influence of the concept on each organization (group).

[Overall Processing of Analysis / Evaluation Unit 4]
First, before describing the processing of each component of the analysis / evaluation unit 4, an overview of the processing of the analysis / evaluation unit 4 as a whole will be described.
FIG. 11 is a flowchart showing the overall processing (S10) of the analysis / evaluation unit 4.
In the following, unless otherwise specified, the organization (group) to be evaluated by the analysis / evaluation program 30 (analysis / evaluation unit 4) shown in FIG. As shown by a dotted line in FIG. 1, a specific example is the case of three mailing lists A, B, and C (groups A, B, and C).
Similarly, the analysis / evaluation program 30 (analysis / evaluation unit 4) is obtained from messages sent and received by these mailing lists A, B, and C (groups A, B, and C) and stored in the log DB 304. The target message is the morphological analysis target, the frequency of nouns in the message body in the message log is the target of the analysis, and the propagation of the concept (in this example, the noun in the message log) and its influence The case of evaluation is a specific example.

Normally, an e-mail message is composed of a header and a body, and the body is composed of a text, a quote, and an attached file.
The analysis / evaluation unit 4 can also analyze attachments in addition to the body of the message, and can also analyze other parts of speech such as verbs, adjectives, or phrases in addition to nouns, or Other than the part of speech, the unit of analysis can be used.

As shown in FIG. 11, in steps 100 to 104 (S100 to S104), the analysis / evaluation unit 4 performs preprocessing on the message log.
That is, in step 100 (S100), the analysis / evaluation unit 4 reads from the log DB 304 the message log transmitted in the mailing lists A, B, and C (groups A, B, and C) during the period to be evaluated.

Further, in step 102 (S102), the analysis / evaluation unit 4 performs a process of dividing each message included in the message log into units that can be analyzed.
That is, for example, the analysis / evaluation unit 4 divides each message into header information and a message body, and divides the message body into a quote and a body.
In step 104 (S104), the analysis / evaluation unit 4 further performs processing such as morphological analysis on the message body obtained from each message, and divides the body into expression elements (words).

FIG. 12 is a flowchart showing the individual tabulation process (S12) shown in FIG.
In step 12 (S12), the analysis / evaluation unit 4 uses expressions (nouns) included in the expression elements obtained as a result of the division and the number thereof to mailing lists A, B, C (groups A, B, C). Each of them is individually counted (individual counting process).

The individual tabulation process by the analysis / evaluation unit 4 will be described in more detail.
As shown in FIG. 12, in step 120 (S120), the analysis / evaluation unit 4 is included in the message log of the mailing list (group) to be processed and obtained from the message main body to be processed. The total number (total number of expressions) of the expressed expressions (nouns (hereinafter also referred to as group k expressions)) is totaled.
In step 122 (S122), the analysis / evaluation unit 4 totals the number of expressions (number of expressions) included in the expression of the group k.

In step 124 (S124), the analysis / evaluation unit 4 counts how many times each of these expressions appears in the message log of group k (number of appearances) for each type of expression of group k.
In step 126 (S126), the analysis / evaluation unit 4 determines whether or not the processing of S120 to S124 has been completed for the expressions obtained from all the messages of the group k.
The analysis / evaluation unit 4 proceeds to the process of S128 when the processes of S120 to S124 are completed for the expressions obtained from all the message bodies of the group k.
In other cases, the analysis / evaluation unit 4 returns to the processing of S120 with the message body that is not yet the processing target of S120 to S124 as the processing target.

In step 128 (S128), the analysis / evaluation unit 4 extracts, as a feature expression, an expression having a large number of appearances from expressions obtained from all the message bodies of the group k.
For example, the analysis / evaluation unit 4 extracts feature expressions by selecting five expressions in descending order of the number of appearances.
In step 130 (S130), the analysis / evaluation unit 4 determines whether or not the processing of S120 to S128 has been completed for all of the mailing lists A, B, and C (groups A, B, and C).

The analysis / evaluation unit 4 proceeds to the totaling section determination process (S14; FIG. 11) when the processes of S120 to S128 are completed for all of the mailing lists A, B, and C (groups A, B, and C).
In other cases, the analysis / evaluation unit 4 proceeds to the processing of S120 with any of the mailing lists (groups) not yet subjected to the processing of S120 to S128 as the processing target.

FIG. 13 to FIG. 15 are diagrams illustrating feature expression lists obtained by summing expressions (nouns) individually for mailing lists A, B, and C (groups A, B, and C).
The analysis / evaluation unit 4 does not simply select the highest number of appearances depending on the content and method of the evaluation, but selects, for example, proper nouns or proper names with a large number of appearances. Or the feature expression may be selected.
By the individual counting process (S12) of the analysis / evaluation unit 4 described above, the mailing lists A, B, and C (groups A, B, and C) are individually included in the message body as shown in FIGS. Expressions (nouns; customers, women, discussions, strategies, technologies, customers, businesses, sites, requests, servers, everyone, ideas, costs, circles, robots, prints, production) and the number of occurrences of these expressions (nouns) Are aggregated to create a list of expressions.
Further, by the individual tabulation process (S12) of the analysis / evaluation unit 4 described above, the mailing lists A, B, and C (groups A, B, and C) among the expressions (nouns) shown in FIGS. Then, an expression having a large number of appearances is selected so as to indicate the list item without a background, and a list of feature expressions is created.

FIG. 16 is a flowchart showing the totaling section determination process (S14) shown in FIG.
FIG. 17 is a flowchart showing the duplicated expression extraction process (S16) shown in FIG.
The analysis / evaluation unit 4 changes the feature expression of each of the mailing lists A, B, and C (groups A, B, and C) obtained by the individual tabulation process (S12; FIGS. 11, 12 to 15) over time. In order to analyze, the total period determination process (S14) which divides | segments the period used as evaluation object into a section is performed.

As shown in FIG. 16, the analysis / evaluation unit 4 first performs a duplicate expression extraction process (S16).
As shown in FIG. 17, in step 160 (S160), the analysis / evaluation unit 4 performs initial setting for the duplicate expression extraction process.
In step 162 (S162), the analysis / evaluation unit 4 refers to the feature expression list for each of the mailing lists A, B, and C (groups A, B, and C) shown in FIGS. It is determined whether there are duplicate feature expressions between the lists.

In other words, the analysis / evaluation unit 4 uses any one of the feature expression lists (FIGS. 13 to 15) of the mailing list (group) to be processed as the other mailing list ( Group) is included in the feature expression list.
The analysis / evaluation unit 4 proceeds to the process of S170 when a duplicated feature expression is included between the feature expression lists of the mailing lists A, B, and C (groups A, B, and C).
In other cases, the analysis / evaluation unit 4 proceeds to the process of S164.

In step 164 (S164), the analysis / evaluation unit 4 determines in the processing of S162 that the feature expression is included between the feature expression lists of the mailing lists A, B, and C (groups A, B, and C). Is a feature expression (new expression) that is newly determined to be duplicated between the feature expression lists of the mailing lists A, B, and C (groups A, B, and C).
In the process of S162, the analysis / evaluation unit 4 determines that the feature expression determined to be included between the feature expression lists of the mailing lists A, B, and C (groups A, B, and C) is a new expression. The process proceeds to S166.
In other cases, the analysis / evaluation unit 4 proceeds to the process of S168.

FIG. 18 is a diagram illustrating a candidate list being created by the duplicate expression extraction process (S16).
In step 166 (S166), the analysis / evaluation unit 4 overlaps between the feature expression lists (FIGS. 13 to 15) of the mailing lists A, B, and C (groups A, B, and C) as shown in FIG. New items are added to a list (candidate list) of feature expressions (candidate expressions) that are included in the feature expressions (duplicate expressions) included in the list.
FIG. 18 illustrates a case where “technology” is added as a new item as a duplicate expression, and “30” is registered as data on the number of appearances in the mailing list A (group A). .

In step 168 (S168), the analysis / evaluation unit 4 determines which of the mailing lists A, B, and C (groups A, B, and C) that are the processing target of the duplicate expression that is determined not to be a new expression in the process of S164. The number of occurrences in the kana is added as data to the duplicate expression item determined not to be a new expression in the process of S164 of the candidate list.
In step 170 (S170), the analysis / evaluation unit 4 increments a target number indicating an item in the feature expression list (FIGS. 13 to 15) of the mailing list (group) to be processed.

In step 172 (S172), the analysis / evaluation unit 4 completes the processes of S160 to S168 for all items in the feature expression list (FIGS. 13 to 15) of the mailing list (group) to be processed. Determine whether or not.
The analysis / evaluation unit 4 proceeds to the process of S174 when the processes of S162 to S168 are completed for all items in the feature expression list (FIGS. 13 to 15) of the mailing list (group) to be processed.
In other cases, the analysis / evaluation unit 4 sets items that have not been processed in S162 to S168 as processing targets, and returns to the processing in S162.

In step 174 (S174), the analysis / evaluation unit 4 completes the processes of S162 to S168 for all items in the feature expression list (FIGS. 13 to 15) of all mailing lists (groups) to be processed. Determine whether or not.
The analysis / evaluation unit 4 completes the processes of S162 to S168 for all items in the feature expression list (FIGS. 13 to 15) of all mailing lists (groups) to be processed. If so, the process proceeds to the processing for determining the number of duplicate expressions (FIGS. 16 and 20).
In other cases, the analysis / evaluation unit 4 processes the first item in any feature expression list (FIGS. 13 to 15) of any mailing list (group) that has not yet been processed in S162 to S168. Then, the process returns to S162.

FIG. 19 is a diagram illustrating a candidate list obtained by the processes of FIGS. 17 and 18.
The candidate list shown in FIG. 19 is obtained by the duplicate expression extraction processing by the analysis / evaluation unit 4 described above.
However, at this stage, the number of divisions and flag items in the candidate list are blank.

FIG. 20 is a flowchart showing the duplicated expression number determination process (S18) shown in FIG.
As shown in FIG. 20, in step 180 (S 180), the analysis / evaluation unit 4 performs initial setting of the overlapping expression number determination process.
In step 182 (S182), the analysis / evaluation unit 4 determines whether there is a candidate expression (next candidate expression) in the candidate list (FIG. 19) that has not yet been processed in S184 to S196.
When there is a next candidate expression, the analysis / evaluation unit 4 proceeds to the process of S184 with any one of the next candidate expressions as a processing target. Otherwise, the analysis / evaluation unit 4 performs the process of contingency table aggregation (S22; FIGS. 11 and 21). move on.

In step 184 (S184), the analysis / evaluation unit 4 obtains the minimum value of the number of appearances of the candidate expression that is the processing target.
In step 186 (S186), the analysis / evaluation unit 4 determines whether or not the number of appearances obtained by the process of S184 is less than a predetermined number (for example, 20).
The analysis / evaluation unit 4 proceeds to the process of S188 when the number of appearances obtained by the process of S184 is equal to or less than a predetermined number (for example, 20), and proceeds to the process of S194 otherwise.

In step 188 (S188), the analysis / evaluation unit 4 selects mailing lists A, B, and C (groups A, B, and C) corresponding to the number of appearances determined to be less than or equal to the predetermined number in the process of S186 from the candidate list. ) Is excluded from the evaluation.
In step 190 (S190), the analysis / evaluation unit 4 determines whether or not the number of mailing lists (groups) to be evaluated is less than 2 in the candidate expression to be processed.
The analysis / evaluation unit 4 proceeds to the process of S192 when the number of mailing lists (groups) to be evaluated is less than 2 in the candidate expression to be processed, and otherwise performs the process of S184. Return.

In step 192 (S192), the analysis / evaluation unit 4 excludes the candidate expression that is the target of processing from the evaluation target, and returns to the processing of S182.
In step 194 (S194), the analysis / evaluation unit 4 equally divides the period to be evaluated, and the number of divisions for providing an evaluation section (the length of the section = the evaluation target). Period length / number of divisions).
The number of divisions is determined based on the approximate condition np ≧ 10 (np is the expected frequency) for the χ ² distribution when performing the χ ² fitness test, and the total number of representations included in the section i obtained by the division the n _i, expressed e _j average occurrence probability p _ei, the total number expressed as n, is calculated by the following equation 1.

In step 196 (S196), the analysis / evaluation unit 4 causes the analysis / evaluation unit 4 to have a minimum number of appearances of a predetermined number (for example, 20) or more in the candidate list by the process of S186, as shown in FIG. In order to indicate that the candidate expression determined to be the object of evaluation, the evaluation object flag is set to 1, the number of divisions calculated by the process of S194 is registered for this candidate expression, and the process returns to S182.
In addition, as shown in FIG. 19 as items without background, all of the mailing lists A, B, and C (groups A, B, and C) are evaluated for “discussion” and “technology” in the candidate expressions. An example is shown in which mailing list B (group B) is an evaluation target for “site” and “server”, and mailing list C (group C) is an evaluation target for “cost”.

Since the candidate expressions “site”, “server”, and “cost” have one mailing list (group) to be evaluated, they are excluded from the evaluation target by the processing of S190.
However, regarding these candidate expressions, if there are two mailing lists (groups) to be evaluated, they are not excluded from the evaluation by the process of S190.

FIG. 21 is a flowchart showing the contingency table aggregation process (S22) shown in FIG.
As shown in FIG. 21, in step 220 (S220), the analysis / evaluation unit 4 performs initial setting of the contingency table aggregation process.
In step 222 (S222), the analysis / evaluation unit 4 has a candidate expression (next candidate expression) in the candidate list (FIG. 19) that has not yet been subjected to the processing of S224 to S236. Determine whether or not.
If there is a next candidate expression, the analysis / evaluation unit 4 proceeds to the process of S224 with any of the next candidate expressions as a processing target, and proceeds to the frequency change determination process (S26; FIGS. 11 and 25) otherwise.

In step 224 (S224), the analysis / evaluation unit 4 refers to the candidate list (FIG. 19), and sets the evaluation target flag of the candidate expression that is the processing target of the mailing list (group) that is the processing target. It is determined whether or not 1.
The analysis / evaluation unit 4 proceeds to the processing of S226 when the evaluation target flag of any candidate expression of the mailing list (group) that is the target of processing is 1.
In other cases, the analysis / evaluation unit 4 processes candidate expressions that are any of the mailing lists (groups) that are the targets of processing and that are not yet the targets of the processing of S226 to S234. Then, the process returns to S222.

In step 226 (S226), the analysis / evaluation unit 4 determines whether the candidate expression that is the target of processing of the mailing list (group) that is the target of processing can be divided.
For example, in the candidate list (FIG. 19), the analysis / evaluation unit 4 can divide the candidate expression when the number of candidate expressions in a mailing list (group) reaches a predetermined number (for example, the number of divisions × 10). Otherwise, it is determined that the candidate expression cannot be divided.

The analysis / evaluation unit 4 proceeds to the processing of S228 when the candidate expression that is the processing target of the mailing list (group) that is the processing target can be divided, and otherwise proceeds to the processing of S236. .
Note that candidate expressions determined to be indivisible in the process of S226 are excluded from the evaluation target and ignored.
Since the expression to be evaluated is determined by the processing so far, the expression determined as the evaluation target is hereinafter also referred to as a confirmed expression.

22 to 24 are diagrams illustrating examples of totaling results of the mailing lists A, B, and C (groups A, B, and C) obtained by the contingency table totaling process (S22) illustrated in FIGS. .
In step 228 (S228), the analysis / evaluation unit 4 divides the evaluation target period according to the number of divisions, and sets a totaling section as shown in FIGS.
Examples of the method for setting the interval for aggregation include a method of equally dividing the evaluation target period according to the number of divisions, or a method of dividing the evaluation period so that the number of messages is equalized for each interval. can do.
For example, in FIGS. 22 to 24, the analysis / evaluation unit 4 equally divides the evaluation target period (30 days) according to the division number 3, and sets three totaling sections 1 to 3 for 10 days each. The case is given as a specific example.

In step 230 (S230), the analysis / evaluation unit 4 counts the number of appearances of each defined expression in the section to be processed.
In step 232 (S232), the analysis / evaluation unit 4 adds up the total number of defined expressions in the section to be processed.
As shown in FIGS. 22 to 24, the analysis / evaluation unit 4 may further count the number of mails in the section targeted for processing.

In step 234 (S234), the analysis / evaluation unit 4 determines whether or not the aggregation has been completed for all the sections set for the mailing list (group) to be processed.
The analysis / evaluation unit 4 proceeds to the process of S236 when the aggregation is completed for all sections set for the mailing list (group) to be processed.
In other cases, the analysis / evaluation unit 4 selects any section that has not yet been processed in S230 and S232 among the sections set for the mailing list (group) that is the target of processing. Return to the processing of S230.

In step 236 (S236), the analysis / evaluation unit 4 determines whether or not the tabulation processing of the definite expression that is the processing target in the processing of S222 has been completed for all mailing lists (groups).
For all mailing lists (groups), the analysis / evaluation unit 4 proceeds to the processing of S222 when the aggregation processing of the defined expressions that are the processing targets in the processing of S222 is completed, and proceeds to the processing of S226 otherwise.
By the split table tabulation processing (S22) described above, the tabulation results shown in FIGS. 22 to 24 are obtained.

FIG. 25 is a flowchart showing the frequency change determination process (S26) shown in FIG.
In the frequency change determination process (S26) shown in FIG. 25, a change (frequency change) in the number of appearances of the definitive expression in each mailing list (group) in each section is detected from the aggregation results shown in FIGS. It is determined which definite expression (concept) has propagated from which mailing list (group) to any other mailing list (group).

As shown in FIG. 25, in step 260 (S260), the analysis / evaluation unit 4 performs initial setting of the frequency change determination process.
In step 262 (S262), the analysis / evaluation unit 4 indicates that the analysis / evaluation unit 4 still includes the results of S264-S274 in the tabulation results (FIGS. 22-24) of each mailing list (group) that is the object of processing. It is determined whether there is a definite expression (next candidate expression) that is not a target of processing.
When there is a next candidate expression, the analysis / evaluation unit 4 proceeds to the process of S264 with any of the next candidate expressions as a processing target, and proceeds to the target scale calculation process (S30; FIGS. 11 and 27) otherwise.

In step 264 (S264), the analysis / evaluation unit 4 initializes the mailing list (group) to be processed (for example, the group A is the processing target).
In step 266 (S266), the analysis / evaluation unit 4 sets the section to be processed as the first section (1).

In step 268 (S268), the analysis / evaluation unit 4 evaluates the change in the appearance frequency of the definitive expression in the evaluation target section of the evaluation target mailing list (group).
In this evaluation, the probability that a definite expression appears (normal appearance probability) when it is assumed that there is no influence from outside throughout the entire period is p _ej, and the total number of expressions in the section to be determined is n _i. And the discriminant shown in the following formula 2 is used, where χ _ie is the number of appearances of the defined expression in the section to be evaluated, and c is the χ ² value of the specified risk factor with a degree of freedom 1 as the discriminant coefficient.
When the value of this discriminant is 0 or more, it is evaluated that a change has occurred in the appearance frequency of the definitive expression. When the value of this discriminant is less than 0, it is evaluated that no change has occurred in the appearance frequency of the definitive expression.

Discriminant shown in the above equation 2, the formula for obtaining the chi ² fit in performing chi ² goodness of fit test (the following formulas 3), as expected values of _t iej = n _i p _ej, as shown below Derived.

  By changing Equation 3, the following Equation 4 is obtained.

  Further, Equation 4 is transformed to obtain Equation 5 below.

Equation 5 can be transformed into Equation 6 below because only one side is sufficient for discrimination when the appearance frequency increases.
By arranging the right side of the following expression 6-1 to 0, the above expression 2 is obtained.
In order to investigate, for example, the situation in which the use of a certain expression is inappropriate and the frequency of appearance decreases due to the influence from other groups (mailing lists), Expression 6-2 may be used as a discriminant. Absent.

In this discriminant (Formula 2), the value of c is 3.84 when the risk factor is 5%, and the value of c is 6.63 when the risk factor is 1%.
In the process of S246, the risk factor may be a value of both 1% and 5% as well as 1% or 5%.
Alternatively, the value of c in the discriminant may be a value corresponding to a risk rate other than statistically general 5% and 1%.
Here, a specific example is given of a case where the analysis / evaluation unit 4 evaluates the frequency change of the definite expression with the values of c corresponding to the risk rates of 5% and 1%, respectively.

  When the discriminant (formula 2) is applied to the definite expression “technique” of section 1 of the mailing list A (group A) shown in FIG. 22, the evaluation result shown in the following formula 7 is obtained.

  Similarly, when the discriminant (formula 2) is applied to the definite expression “technique” of section 3 of the mailing list A (group A) shown in FIG. 22, the evaluation result shown in the following formula 8 is obtained.

In step 270 (S270), the analysis / evaluation unit 4 determines whether or not the value of the discriminant (expression 2) calculated in S268 is less than zero.
The analysis / evaluation unit 4 proceeds to the process of S274 when the value of the discriminant (expression 2) calculated in S268 is less than 0, and otherwise proceeds to the process of S272.

FIG. 26 is a diagram illustrating a frequency change flag table created by the frequency change determination process (S26) shown in FIGS.
In FIG. 26, “*” indicates a frequency flag when c is set to a value corresponding to a risk factor of 5% in the discriminant, and “**” indicates that c is a risk factor of 5%. Indicates the frequency flag when set to the corresponding value.
In step 272 (S272), as shown in FIG. 26, the analysis / evaluation unit 4 determines the expressions to be evaluated (“technology” / “discussion”), mailing lists A, B, C (group A, B, C) and a frequency change table combining sections 1 to 3 are created, and a frequency change flag is set to an item in which a frequency change is detected by the processing of S270 among the items of the frequency change table.

In step 274 (S274), the analysis / evaluation unit 4 determines whether or not the evaluation of the change in the appearance frequency of all the sections has been completed for the fixed expression and the mailing list (group) that are to be evaluated.
The analysis / evaluation unit 4 proceeds to the process of S276 when the evaluation of the change in the appearance frequency of all the sections is completed for the confirmed expression and the mailing list (group) that are the objects of the evaluation.
In other cases, the analysis / evaluation unit 4 sets, as processing targets, sections that have not yet been evaluated for appearance frequency changes in the defined expressions and mailing lists (groups) that are targeted for evaluation. Return to processing.

In step 276 (S276), the analysis / evaluation unit 4 determines whether or not the evaluation of the change in the appearance frequency with respect to the defined expression that is the target of the evaluation has been completed for all mailing lists (groups).
The analysis / evaluation unit 4 proceeds to the processing of S262 when the evaluation of the change in the appearance frequency with respect to the fixed expression to be evaluated is completed for all the mailing lists (groups).
In other cases, the analysis / evaluation unit 4 sets a mailing list (group) that has not yet been evaluated for the change in the appearance frequency for the defined expression that is the target of evaluation, and proceeds to the process of S266.

Through the above-described frequency change determination process, each item in the frequency change table shown in FIG. 26, that is, a fixed expression, an evaluation target unit (mailing lists A, B, and C (groups A, B, and C)) and an evaluation section are obtained. For each combined item, the change in the appearance frequency of each defined expression is evaluated, and the evaluation result is set in the frequency change table.
By analyzing this frequency change table, as indicated by an arrow in FIG. 26, it is evaluated which route the confirmed expression has propagated from which mailing list (group) to which other mailing list at which time. can do.

For example, referring to FIG. 26, the evaluation candidate “technology” is propagated from the mailing list C (group C) to the mailing list B (group B) during the sections 1 to 2, and further, during the sections 2 to 3, It can be seen that the message was propagated from mailing list C (group C) or mailing list B (group B) to mailing list A (group A).
Further, FIG. 26 shows that the evaluation candidate “discussion” is propagated from the mailing list B (group B) to the mailing list C (group C) during the sections 1 to 2.

FIG. 27 is a flowchart showing the target scale calculation process (S30) shown in FIG.
FIG. 28 is a flowchart showing the participant list creation processing (S32) shown in FIG.
As shown in FIG. 27, the analysis / evaluation unit 4 performs a participant list creation process in step 32 (S32).
In this participant list creation process, personal information of members (members) of each mailing list (group) from the organization / personal DB 310 (FIG. 6) (FIG. 8; organizational information of each mailing list (group) as necessary) 7)) is read out and a list of participating groups is generated.
As shown in FIG. 28, in step 320 (S320), the analysis / evaluation unit 4 performs initial setting of the participant list creation process.

In step 322 (S322), the analysis / evaluation unit 4 determines whether there is a mailing list (group) that has not yet been processed in S224 to S330.
The analysis / evaluation unit 4 proceeds to the process of S324 with any mailing list (group) as a process target when there is a mailing list (group) that is not yet the process of S324 to S330. Proceed to the counting process (S36; FIGS. 27 and 34).

FIG. 29 is a diagram illustrating personal IDs of members of mailing lists A, B, and C (groups A, B, and C).
For example, members with personal IDs (FIG. 8) shown in FIG. 29 belong to mailing lists A, B, and C (groups A, B, and C).
In step 324 (S324), the analysis / evaluation unit 4 still performs the processing of S326 to 330 from the organization / individual DB 310 in the personal information (FIG. 8) of the members of the mailing list (group) to be processed. Any one that is not the target is read, and the personal ID (member ID) is taken out.

In step 326 (S326), the analysis / evaluation unit 4 determines whether any of the personal IDs obtained by the processing of S324 has become a processing target until then. It is determined whether or not it is a personal ID.
The analysis / evaluation unit 4 determines that if any of the personal IDs obtained by the processing of S324 is a new personal ID that has not been processed before, Proceed to the process, otherwise proceed to the process of S330.

FIG. 30 is a diagram illustrating a participant list created by the participant list generation process (S32; FIGS. 27 and 28).
In step 328 (S328), the analysis / evaluation unit 4 adds an item for a new personal ID to the participant list shown in FIG.
In step 330 (S330), the analysis / evaluation unit 4 adds the ID (A, B, or C) of the mailing list (group) to which the member belongs to the item of the personal ID to be processed in the participant list. To do.

In step 332 (S332), the analysis / evaluation unit 4 determines whether or not the processing has been completed for all members of the mailing list to be processed.
The analysis / evaluation unit 4 proceeds to the process of S322 when the process is completed for all members of the mailing list to be processed, and proceeds to the process of S324 otherwise.
As a result of the above processing, the analysis / evaluation unit 4 has each member belong to each personal ID of the member belonging to any of the mailing lists A, B, C (groups A, B, C) as shown in FIG. A participant list in which IDs (A, B, C) of mailing lists (groups) to be associated are created is created.

FIG. 31 is a diagram illustrating members who belong to the mailing lists A, B, and C (groups A, B, and C) redundantly.
FIG. 32 is a diagram illustrating categories of members belonging to the mailing lists A, B, and C (groups A, B, and C) in duplicate.
FIG. 33 is a diagram illustrating a classification table obtained by the category tabulation process (S36; FIGS. 27 and 34).
As shown in FIG. 27, in step 36 (S36), the analysis / evaluation unit 4 performs category aggregation processing.

First, an overview of the category tabulation process (S36; FIGS. 27 and 34) will be described.
As shown in FIG. 31, members belonging to any one of mailing lists A, B, and C (groups A, B, and C) and any two of mailing lists A, B, and C (groups A, B, and C) There may be more than one member belonging to more than one.
For example, in FIG. 31, there are six members who belong only to mailing list A (group A), one member who belongs to both mailing lists A and B (groups A and B), and mailing lists A, B, and C (groups). A case where there is one member belonging to all of A, B, C) is shown.

Each area shown in FIG. 31 can be used as a category for aggregation according to the overlapping state of mailing lists A, B, and C (groups A, B, and C). As shown in FIG. The names a, b, c, b _{a to} c _b , c _ab are given.
By the category totaling process, the individual IDs of the members belonging to the respective categories shown in FIG. 32 and the number of members belonging to the respective categories are totaled to obtain the classification table shown in FIG.
Note that category a, b _a is represented by the following formula 9.

FIG. 34 is a flowchart showing the category totaling process (S36) shown in FIG.
The category tabulation process (S36) creates an empty table with columns for entering ID and number of people in (2 ^{2 groups-1} ) rows at the initial setting, and adds data to the categories to which each member belongs. Thus, the classification table shown in FIG. 33 is created.
As shown in FIG. 34, in step 360 (S360), the analysis / evaluation unit 4 initializes the classification table shown in FIG.
In step 362 (S362), the analysis / evaluation unit 4 reads items of the participant list one by one.

In step 364 (S364), the analysis / evaluation unit 4 calculates a classification number i from the items in the participant list read out in the process of S362.
The classification number i is set to g as the group ID of the mailing list (group) to which the read member belongs when the group IDs 0 to 2 are assigned to the mailing lists A, B, and C (groups A, B, and C). , Calculated from Equation 10 below.
For example, from the group ID = 0 of the mailing list A (group A) and the group ID = 2 of the mailing list C (group C), the classification number (category number) i of the member of the personal ID 7 shown in FIG. Calculated.

In step 366 (S366), the analysis / evaluation unit 4 increments the number of items of the classification number i obtained by the process of S362.
In step 368 (S368), the analysis / evaluation unit 4 determines whether or not processing has been completed for all members of the participant list.
The analysis / evaluation unit 4 proceeds to the process of the direct influence calculation process (S110) shown in FIG. 11 when the process is completed for all members of the participant list, and proceeds to the process of S362 otherwise. .

As shown in FIG. 11, in step 110 (S110), the analysis / evaluation unit 4 calculates the direct influence degree.
The degree of influence is defined as follows.
Determined expression e _j to the message in the mailing list _(group), x iej number appeared, the probability of occurrence of definite expression e _j assumed not to be affected from other mailing list (group) (usually appearance probability) p and _ei, and then fully express the number in the interval i of that list (group) and n _i, for the appearance of definite expressions e _j in the list (group), the degree of apparent influence of other mailing list (group) ( Apparent external influence) is defined as x _iej / n _i .

As shown in the following equation 11, is a mailing list (group) itself, a minute degree of had the appearance probability p _ei deterministic expression e _j effect (internal impact), using the coefficient k, external influences the apparent By correcting from the degree, the actual external influence degree can be obtained.
That is, in some mailing list (group), the number of the defined expression e _j appearing without influence of other mailing list (group) and n _i p _ej, multiplied by a factor k, subtracting from outside influence apparent Thus, the actual external influence degree is obtained.

In Expression 11, when the appearance probability p _ej is the maximum value (1), since the influence from other mailing lists (groups) is 0, Expression 11 can be transformed into the following Expression 12.

Effect of all the determined expression e _j is the number of occurrences of the defined expression e _j and x _IEJ, a normal occurrence probability as p _ej, represented by the following formula 13.
The degree of influence expressed by Expression 13 indicates the degree of influence that all definite expressions subjected to evaluation give to a certain mailing list (group) via another mailing list (group).

Here, as can be seen with reference to FIGS. 31 and 32, the degree of influence between mailing lists (groups) is the size of each of the mailing lists A, B, and C (groups A, B, and C) and a plurality of mailing lists (groups). ) Needs to be compensated for the effects that members (shared members) belong to.
Therefore, the influence degree is corrected as follows so that the magnitude of the influence degree between the affected mailing lists (groups) can be determined in consideration of the shared member and the scale of the affected mailing list (group).

First, the p _ej the probability of occurrence of definite expressions e _j in list B (Group B), impact I _AB for mailing B (Group B) mailing list A (Group A) was judged frequency is changed The number of appearances of the defined expression ej in the section is x ′ _Bej, and the total number of expressions in the section in which it is determined that the appearance frequency has changed is N ′ _B.

When the degree of influence I _AB shown in Equation 14 is corrected using the direct correction coefficient α with the number of people included in the categories a and b _a shown in FIGS. 32 and 33 as a and b _a , the degree of influence after correction is corrected. I _AB is expressed as shown in Equation 15.
It should be noted that this direct correction coefficient α, the category b _a, mailing list A, B (group A, B) because it belongs to both, members belonging to the category b _a is, the influence on the mailing list B (Group B) Used for underestimation.

The impact on mailing lists with many people is considered to be a significant impact.
However, even if there are only a large number of members, it cannot be said that the influence level is high on a mailing list where mail does not actually flow.
Therefore, it is assumed that how many people the mail (message) will flow to how many people is the scale, and the correction can be performed in consideration of the scale.
That is, the correction is performed in consideration of the frequency of message delivery as shown in Expression 16, in addition to the influence degree I _AB after correction shown in Expression 15.
If the total number of messages of members belonging to category b is M _b and the length of the period to be evaluated is T _b , the size of the mailing list (group) to be evaluated is the number of messages delivered per unit period (M _b / T _b )

For example, the mailing list is obtained by performing correction similar to the correction shown in Expression 16 on the aggregation result of the mailing list A (group A) shown in FIG. 22 and the data shown in the classification table shown in FIG. When the degree of influence I _CA on mailing list A (group A) is calculated from C (group C), the result shown in the following equation 17 is obtained when the direct correction coefficient α = 0.5.

Similarly, when the degree of influence I _BA from the mailing list B (group B) to the mailing list A (group A) is calculated, the result shown in the following equation 18 is obtained.

Further, for example, the same correction as that shown in Expression 16 is performed on the aggregation result of the mailing list B (group B) shown in FIG. 23 and the data shown in the classification table shown in FIG. When the degree of influence I _CB on the mailing list B (group B) is calculated from the mailing list C (group C), the result shown in the following equation 19 is obtained.

As shown in FIG. 11, in step 112 (S112), the analysis / evaluation unit 4 performs correction for reflecting an indirect influence on the influence degree calculated by the process of S110.
A certain expression not only affects one mailing list (group) from another mailing list (group), but also, for example, certain expressions affect mailing list A (group A) to mailing list B (group B). In addition, the same expression may affect mailing list B (group B) to mailing list C (group C).

That is, the mailing list C (group C) does not receive the influence of a certain expression directly from the mailing list A (group A), but via the mailing list B (group B), the mailing list A (group A)-> mailing list B. It is received indirectly through a route of (Group B)-> Mailing list C (Group C).
The degree of influence I _AB that all the defined expressions give to mailing list B (group B) from mailing list A (group A) is calculated as in equation 20 below, as in equation 17 and the like.

Similarly, the degree of influence I _BC that all the defined expressions have on the mailing list C (group C) from the mailing list B (group B) is calculated as in the following equation (21).

The degree of influence I _ABC indirectly given to all the defined expressions from the mailing list A (group A) to the mailing list C (group C) via the mailing list B (group B) is calculated as shown in the following expression 22. .

When the degree of influence I _ABC (formula 22) is subtracted from the degree of influence I _BC (formula 21) as shown in the following formula 23, all the definite expressions are indirectly sent from the mailing list A (group A) to the mailing list C (group). The degree of influence on the mailing list C (group C) can be obtained from the mailing list B (group B) excluding the degree of influence on C).

A specific example is shown.
For example, from the tabulation results shown in FIGS. 22 to 24 and the classification table shown in FIG. 33, all definite expressions are sent from the mailing list C (group C) to the mailing list A (group B). The degree of influence I _CBA on group A) is calculated as in the following equation 24, with the indirect correction coefficient β = 0.2.

When the degree of influence I _CBA (formula 24) is subtracted from the degree of influence I _BA , as shown in the following formula 25, the influence that all definite expressions are indirectly given to the mailing list A (group A) from the mailing list C (group C) The degree of influence on the mailing list A (group A) can be obtained from the mailing list B (group B) excluding the degree.

From these calculation results, the degree of influence that all the defined expressions have on the other mailing lists (groups) from the mailing list C (group C) is calculated as in the following Expression 26.
Further, the degree of influence that all the defined expressions have on the other mailing lists (groups) from the mailing list B (group B) is calculated as in the following Expression 27.

  As a result of the above processing, the analysis / evaluation unit 4 has an influence that a certain expression gives to another mailing list (group) from a certain mailing list (group), and thus influences another mailing list (group) on another mailing list (group). The degree is evaluated with high accuracy.

[Processing of components of analysis / evaluation unit 4]
Refer to FIG. 10 again.
The evaluation target / range specifying unit 400 receives a user operation on the display / input device 206 (FIG. 2) via the UI unit 312 (FIG. 6), and the evaluation target unit (mailing list in the examples so far), and Specify the expression to be evaluated (noun in the above examples).
Similarly, the evaluation target / range designation unit 400 designates a period to be dealt with in the evaluation.

The member classification unit 402 performs the participant list creation process shown in FIGS.
The preprocessing unit 420 performs the processing (preprocessing) of S100 to S104 illustrated in FIG.
The aggregation unit 422 performs the individual aggregation process (S12) shown in FIGS. 11 to 15, the contingency table aggregation shown in FIGS. 11, 21 to 24, and the category aggregation shown in FIGS. 27 and 31 to 34. Processing (S36) is performed.

The feature extraction unit 424 performs a feature expression extraction process included in the individual tabulation process (S12) shown in FIGS.
The section determination unit 426 performs the total section determination process (S14) illustrated in FIGS. 11 and 16 to 20.
The influence degree calculation unit 428 is set in advance with the frequency change determination process (S26) shown in FIGS. 11, 25, and 26, the target scale calculation process (S30) shown in FIGS. The direct influence calculation processing (S110) and indirect influence correction processing (S112) shown in FIG. 11 are performed using the correction coefficients α and β.

The result management unit 440 stores and manages the evaluation result obtained by the influence calculation unit 428 in the analysis / evaluation DB 308.
Further, the result management unit 440 receives a user operation on the display / input device 206 (FIG. 2) via the UI unit 312 (FIG. 6), and displays or outputs the evaluation result stored in the analysis / evaluation DB 308. .

[Overall operation of network system 1]
Hereinafter, the overall operation of the network system 1 (FIG. 1) will be described.
Organizations on the network system 1 (for example, mailing lists A, B, and C assembled in the network system 1) and members of these mailing lists A, B, and C transmit messages by electronic mail.
The analysis / organization apparatus 3 stores a message transmitted on the mailing lists A, B, and C in the log DB 304 as a message log (FIG. 9; organization communication information).

The user of the analysis / organization apparatus 3 activates the analysis / evaluation program 30 (FIGS. 6 and 10), the concept (words, etc.) to be evaluated, the period and organization to be evaluated (mailing lists A, B, C) or the like is designated, and processing for evaluating the influence of expression is executed.
The analysis / evaluation unit 4 (FIG. 10) of the analysis / evaluation program 30 reads the message log from the log DB 304 and, as shown in FIGS. 11 to 34, the characteristic expressions (determined expressions) are mailing lists A and B. , C, which is affected, and the degree of influence of characteristic expressions (determined expressions).
The analysis / evaluation unit 4 stores and manages the evaluation result in the analysis / evaluation DB 308, and displays or outputs the evaluation result in accordance with a user operation.

[Second Embodiment]
In the first embodiment of the present invention, as shown with reference to Equation 15 and the like, it is directly corrected as a coefficient for correcting the degree of influence I _AB from the mailing list A (group A) to the mailing list B (group B). The coefficient α is used.
Similarly, as shown with reference to Expression 22 and the like, the influence degree I indirectly given from the mailing list A (group A) to the mailing list C (group C) via the mailing list B (group B) _An indirect correction coefficient β is used as a coefficient for correcting _ABC .
Hereinafter, as a second embodiment of the present invention, a second analysis / evaluation program 50 (FIGS. 35 and 36) executed in the analysis / evaluation apparatus 3 in place of the first analysis / evaluation program 30 will be described. .

FIG. 35 is a diagram showing a configuration of a second analysis / evaluation program 50 used in the analysis / evaluation apparatus 3 shown in FIG. 1 instead of the first analysis / evaluation program 30 shown in FIG. .
FIG. 36 is a diagram showing a configuration of the second tabulation / analysis unit 52 of the analysis / evaluation program 50 shown in FIG.

As shown in FIG. 35, the second analysis / evaluation program 50 replaces the first aggregation / analysis unit 4 of the first analysis / evaluation program 30 shown in FIG. Further, a configuration is adopted in which a questionnaire survey unit 340, a questionnaire result DB 342, and a correction coefficient calculation unit 302 are added.
Further, the second tabulation / analysis unit 52 employs a configuration in which a correction processing unit 520 is added to the first tabulation / analysis unit 4 shown in FIG.

The analysis / evaluation program 50 uses these components to correct the results of the questionnaire survey for the members of the network system 1 and the organization information, personal information, and organization communication information (organization communication information; FIGS. 7 to 9). The coefficients α and β (Equation 15, Equation 22, etc.) are improved.
In addition, the analysis / evaluation program 50 has a direct influence degree between two mailing lists (groups) and an indirect influence between two mailing lists (groups) in which one or more mailing lists (groups) are interposed. The degree of influence is corrected by the calculated correction coefficients α and β so that the influence degree can be obtained more accurately.

[Questionnaire survey and correction factors α, β calculation]
Here, a questionnaire survey by the analysis / evaluation program 50 and a method of calculating the correction coefficients α and β will be described.
First, in two mailing lists A and B (groups A and B) with overlapping members (members), a direct correction coefficient for correcting the degree of influence I _AB from the mailing list A (group A) to the mailing list B (group B) A method of calculating α will be described.

FIG. 37 is a diagram exemplifying two mailing lists A and B (groups A and B) having overlapping members (members).
As shown in FIG. 37, the mailing lists A and B (groups A and B) include members a and b (category a and b) belonging to only one of the mailing lists A and B, and members b belonging to both the mailing lists A and B, respectively. _a (category b _a ).
As already mentioned, the direct correction coefficient alpha, the category b _a, mailing lists A, B (Group A, B) so belongs to both members belonging to the category b _a is the mailing list B (Group B) It is used to estimate the influence to be less, and is used to correct the influence degree I _AB as shown in Equation 28.

However, in Equation 28 (Equation 28 is substantially the same as Equation 16 except that the symbols are different), T _b is an evaluation period of the influence degree I _AB , and M _b is an e-mail in the evaluation period T _b N _b1 is the total number of concepts (such as nouns in the message log) included in all messages in the evaluation period T _b , and m _xb1 is the number of occurrences of the concept x in the evaluation period T _b P _xb0 is the probability that the concept x appears throughout the entire period not limited to the evaluation period evaluation period T _b in the mailing list B (group B).

FIG. 38 is a diagram showing a method for calculating the direct correction variable α shown in Equation 28.
As shown in FIG. 38, the category b _a a member (members) are impact given from a mailing list A (group), the category b _a a member (members) are averagely from belongs to the mailing list (group) It can be considered that it is equal to the degree of influence received.
In other words, category b _a a member (members) is, degree of influence given from the mailing list A (group), the number of mailing lists category b _a a member (members) were affected (group), the category b _a members the number of groups the members of the correction coefficient _{α 1 (α 1 = (b} a resultant by dividing the number of mailing lists (groups) (member) belongs has been affected) ÷ (b _a group members can belong Given as number)).
Of the two numerical values used in the calculation of the correction coefficient alpha ₁ (average degree of influence received from affiliated mailing), the number of groups the members of b _a is affected, the network system 1 will be described below with respect to members (Member) can be known by performing a survey, the number of groups the members of b _a belongs can be calculated from the personal information, organizational information and organizational communication information shown in FIGS. 7 to 9.

In addition, as shown in FIG. 38, category b _a a member (members), but the effect on the mailing list B (group), category b _a a member (members) is, to speak against belongs to the mailing list (group) It can be considered that it is equal to the average speech level.
In other words, category b _a a member (members), but the effect on the mailing list B (group), the number of mailing lists that category b _a a member (members) has issued an e-mail (Group), category b _a of members ( Correction coefficient α ₂ obtained by dividing by the number of mailing lists (groups) to which the member belongs (average of the proportion of mailing lists that speak in the mailing list to which the members belong; α ₂ = (the number of groups to which members of b _a sent e-mails) ) / (Number of groups to which members of b _a belong)).
Any of the two numerical values used for calculating the correction coefficient α ₂ can be calculated from the personal information, organization information, and organization communication information shown in FIGS.

Impact from a mailing list A (group A) for the mailing list B (Group B), the category b _a a member (members) is, the effect obtained from the mailing list A, since it is considered that due to the fact that on to the mailing list B, formula The direct correction coefficient α shown in 28 and the like is calculated from the correction coefficients α ₁ and α ₂ defined as described above, as shown in Expressions 29-1 and 29-2.

In the above formula, α ₁ × α ₂ gives a ratio of transmitting a topic taken from the mailing list A to another mailing list.
However, in the formula 29-2, m represents the average of the number of categories b _a belonging to members (members) belongs to the mailing list (group).
The correction coefficient α ₂ indicates the degree of transmission of _a member (member) of category ba from _a certain mailing list (group) to another mailing list (group).

Therefore, as shown in Expression 29-2, in order to exclude the degree of utterance for the mailing list (group) of the sender from the correction coefficient α ₂ , the direct correction coefficient α obtained by Expression 29-1 is (m−1 ) / M, a direct correction coefficient α more accurately reflecting the actual situation can be obtained.
Incidentally, the tissue communication information shown in FIG. 9, by more detailed analysis, it is possible to calculate the floor level for each category b _a a member (members), strictly determined for each category b _a a member (member) It is also possible to calculate the correction coefficient α ₂ based on the degree of speech.

Next, a questionnaire survey by the analysis / evaluation program 50 will be described.
FIG. 39 shows a GUI (Graphical User Interface) for a questionnaire displayed on the display / input device 206 (FIG. 2) of the client computer 20 (FIG. 1) by the questionnaire survey unit 340 of the analysis / evaluation program 50 shown in FIG. It is a figure which illustrates an image.
In FIG. 39, the analysis / evaluation program 50 affects each member (member) of the network system 1 from any of the mailing lists A to F (groups A to F) to which each member (member) belongs. The GUI image used when investigating whether or not a plurality of answers is possible is illustrated.

As already described, whether each member (member) of the network system 1 has been affected by which mailing list (group) is affected by, for example, the display / input device 206 of the client computer 20 as shown in FIG. It is possible to know by displaying a questionnaire GUI image and totaling the answers of members (members) to the GUI image as questionnaire survey results.
In addition, in response to the questionnaire survey returned to the analysis / evaluation apparatus 3 from the client computer 20 (FIG. 1) of each member (member), in addition to the response to the questionnaire question (FIG. 39), the organizational communication shown in FIG. It is sufficient if information similar to the information is included.
In addition, FIG. 39 illustrates a GUI image configured so that members (members) of the network system 1 can answer only whether they are affected by each mailing list (group). It is also possible to conduct a questionnaire survey using a GUI image in which members (members) of the network system 1 can answer the degree of influence from each mailing list (group) in, for example, five levels.

FIG. 40 is a diagram showing three mailing lists A, B, and C (groups A, B, and C) having overlapping categories (the same as FIGS. 40 and 32).
Next, from the mailing list A (group A) shown in FIG. 40, the indirect correction coefficient β for correcting the degree of influence I _ABC indirectly given to the mailing list C (group C) via the mailing list B (group B) is calculated. The method will be described.
The indirect correction coefficient β is used together with the direct correction coefficient α to correct the influence degree I _ABC as shown in Expression 30 (Equation 30 is substantially the same as Expression 22 except that the symbol is different).

FIG. 41 is a diagram showing the indirect correction coefficient β shown in Equation 30 and the like.
As shown in FIG. 41, the influence from the mailing list A (group A) to the mailing list B (group B) is 1 because there is no mailing list (group) interposed between them.
However, the influence from mailing list A (group A) to mailing list C (group C) is thought to be attenuated by an indirect correction coefficient β (0 ≦ β ≦ 1) because mailing list B (group B) is interposed between them. It is done.

FIG. 42 is a diagram illustrating mailing lists A to D (groups A to D) that are affected in a state where a plurality of mailing lists (groups) can intervene.
Members of mailing list D (group D) affected by each of mailing lists A and B (groups A and B) shown in FIG. 42 based on the results of questionnaire surveys (members, etc.) for members (members) of network system 1 (Members) can be aggregated.
For example, the ratio of members (members) of the mailing list D (group D) who answered that they were influenced by the mailing lists A to C (groups A to C) is expressed as a correction coefficient β ′ _D, and similarly the mailing lists A, B, D ( The ratio of members (members) of the mailing list C (group C) who answered that they were influenced by the groups A, B, and D) is expressed as a correction coefficient β′C.
The average of the correction coefficients obtained for all groups in this way is defined as β ′.

Also, other mailing lists that cannot be directly affected by a certain mailing list (group) based on questionnaire surveys of members (members) of the network system 1 and the personal information and organization information shown in FIGS. When (group) is affected, the number of mailing lists (groups) interposed between them can be obtained.
For example, the numbers n ′ of mailing lists (groups) interposed between the mailing lists A to C (groups A to C) shown in FIG. 42 are 0, 1, 2, and 1, respectively, and 1 is added to these numbers. If the number of hops is defined as the number of hops (n ′ + 1), the number of hops (n ′ + 1) between the mailing lists A to C (groups A to C) and the mailing list D (group D) is 1, 2, 3 respectively. It becomes.
Therefore, the average value (n + 1) of the three hop numbers (n ′ + 1) between the mailing lists A to C (groups A to C) and the mailing list D (group D) is 2.

In the example shown in FIG. 42, the correction coefficient β ′ calculated for the mailing list D (group D) is an indirect to be defined between the mailing lists B and D (groups B and D) as shown in Equation 31. An average value β of the correction coefficient and the indirect correction coefficient defined between the mailing lists A and C (groups A and C) is set between the mailing list D (group D) and the mailing lists A to C (groups A to C). An average value n of the number of intervening mailing lists (groups), that is, a value β ⁿ that is a power of a value n obtained by subtracting 1 from the average number of hops (n + 1) is shown.
Therefore, as shown in Equation 31, by calculating the nth root of the correction coefficient β ′ calculated for the mailing list D (group D), the average indirect correction coefficient β in the mailing lists A to D (groups A to D) is calculated. (= Β ′ ^{1 / n} ) can be obtained.
As described above, by performing a questionnaire survey and using the personal information, organization information, and questionnaire results shown in FIGS. 7 to 9, the correction coefficients α and β can be calculated. The degree of influence (I _AB , I _ABC ) corrected using β can be calculated.

With respect to the example shown in FIG. 42, the indirect influence degree β calculation method will be described in more detail.
FIG. 43 is a first diagram illustrating a specific example of processing for obtaining the indirect influence degree β from the questionnaire survey in the mailing list (group) illustrated in FIG. 42.
43, members (members) of ID = 5 to 7, 5 to 10, 10 to 16, and 14 to 21 belong to the mailing lists A to D (groups A to D; FIG. 42), respectively. As an example, as shown in Chart b, members (members) with ID = 5-7 belong to both mailing lists A and B (groups A and B), and mailing lists B and C (groups) The members (members) with ID = 10 belong to both B and C), and the members (members) with ID = 14 to 16 belong to both mailing lists C and D (groups C and D).

As shown in diagram b of FIG. 43, flag 1 is added to the relationship between mailing lists (groups) including overlapping members (members) (A∩B, B∩C, and C∩D in diagram b of FIG. 43). When a flag-1 is added to the relationship between mailing lists (groups) that do not include duplicate members (members) and expressed in a matrix format, a matrix c shown in FIG. 43 can be obtained (in the matrix c, The relationship between the same mailing lists (groups) such as A∩A is meaningless and is set to 0).
The matrix c shown in FIG. 43 can be processed as described later with reference to FIG. 44 to obtain a matrix d in which each element indicates the number of hops (n ′ + 1) between mailing lists (groups). (However, as will be described later, in the matrix d, -1 indicates that there is no influence between the mailing lists (groups) indicated by the element).

The correction coefficients α and β are obtained by appropriately executing processing between mailing lists (groups) that overlap in an arbitrary manner in addition to between mailing lists (groups) that overlap each other as shown in FIG. Can be calculated.
Also, when surveying the impact of mailing lists (groups) that you do not belong to, it may be difficult to ask members (members) whether they are affected. You may ask if you know what is being discussed.
The specific indirect influence degree β calculation process by the correction coefficient calculation unit 350 will be described later with reference to FIGS. 45 and 46.

[Analysis / Evaluation Program 50]
In the analysis / evaluation program 50 (FIG. 35), as described above with reference to FIG. 39, the questionnaire survey unit 340 surveys the client computer 20 of each member (member) of the network system 1 (FIG. 1). I do.
The questionnaire survey unit 340 stores the answer result returned from the client computer 20 of each member (member) in the questionnaire result DB 342.

As described above with reference to FIGS. 37, 38, and 40 to 43, the correction coefficient calculation unit 350 performs questionnaire results stored in the questionnaire result DB 342, personal information, organization information, and organization communication information (see FIG. 7 to FIG. 9), the correction coefficients α and β are calculated and output to the correction processing unit 520 (FIG. 36) of the tabulation / analysis unit 52.
With reference to FIG. 44, the hop count calculation process (S40) used for the correction coefficient α and β calculation processes by the correction coefficient calculation unit 350 will be further described.

FIG. 44 is a flowchart showing the hop count calculation processing (S40) by the correction coefficient calculation unit 350 shown in FIG.
As shown in FIG. 44, in step 400 (S400), the correction coefficient calculation unit 350 initializes the parameter i corresponding to the row of the matrix c illustrated in FIG. 43 (i = 1).

In step 402 (S402), the correction coefficient calculation unit 350 initializes the parameter j corresponding to the column of the matrix c illustrated in FIG. 43 (j = 1).
In step 404 (S404), the correction coefficient calculation unit 350 determines whether the element (Lij) in the i-th row and j-th column of the matrix c illustrated in FIG. 43 is greater than 0 (Lij> 0).
The correction coefficient calculation unit 350 proceeds to the process of S406 when the element (Lij) in the i-th row and the j-th column of the matrix c is larger than 0, and otherwise proceeds to the process of S420.
By the processing of S404, among the elements of matrix c, element Lij = 0 (when i = j) and element Lij = -1 (when there is no duplication among members (members) between mailing lists (groups)) It will be ignored from the hop count calculation process.

In step 406 (S406), the correction coefficient calculation unit 350 initializes the parameter k (k = 1).
In step 408 (S408), the correction coefficient calculation unit 350 determines whether the element (Ljk) in the jth row and the kth column of the matrix c illustrated in FIG. 43 is greater than 0 (Ljk> 0).
The correction coefficient calculation unit 350 proceeds to the process of S410 when the element (Ljk) of the jth row and the kth column of the matrix c is greater than 0, and otherwise proceeds to the process of S414.

In step 410 (S410), the correction coefficient calculation unit 350 sets the parameter p to the sum of elements Lij and Ljk (p = Lij + Ljk).
The parameter p indicates the number of hops of the j-th mailing list (group) that is indirectly influenced by the i-th mailing list (group).

In step 412 (S412), the correction coefficient calculation unit 350 determines whether or not the element Ljk is a value from 0 to p.
The correction coefficient calculation unit 350 proceeds to the process of S414 when the element Ljk is a value from 0 to p, and proceeds to the process of S416 otherwise.
When the element Lik is other than 0 to p by the process of S412, that is, when the i-th mailing list (group) does not affect the k-th mailing list (group), or from the i-th mailing list (group). When the number of hops in the mailing list (group) is larger than the value of the parameter p, the parameter p indicating the number of hops is updated.

In step 414 (S414), the correction coefficient calculation unit 350 increments the parameter k (k <−k + 1).
In step 416 (S416), the correction coefficient calculation unit 350 substitutes the value of the parameter p for the elements Lik and Lki.
The process of S416 is performed to set the elements Lik and Lki to the same value (p).

In step 418 (S418), the correction coefficient calculation unit 350 determines whether or not the parameter k is larger than the number of rows and columns N of the matrix c illustrated in FIG. 43 (k> N).
The correction coefficient calculation unit 350 proceeds to the process of S420 when the parameter k is larger than the number of rows / columns N, and proceeds to the process of S408 otherwise.
In step 420 (S420), the correction coefficient calculation unit 350 increments the parameter j (j <−j + 1).

In step 422 (S422), the correction coefficient calculation unit 350 determines whether the parameter j is larger than the number of rows / columns N of the matrix c illustrated in FIG. 43 (j> N).
The correction coefficient calculation unit 350 proceeds to S424 when the parameter j is larger than the number of rows / columns N, and otherwise proceeds to the processing of S404.
In step 424 (S424), the correction coefficient calculation unit 350 increments the parameter i (i <−i + 1).

In step 426 (S426), the correction coefficient calculation unit 350 determines whether or not the parameter i is larger than the number of rows / columns N of the matrix c illustrated in FIG. 43 (j> N).
The correction coefficient calculation unit 350 ends the process when the parameter i is greater than the number of rows / columns N, and otherwise proceeds to the process of S402.
Through the processing of S40 described above, the correction coefficient calculation unit 350 obtains a matrix d indicating the number of hops between the affected mailing lists (groups) from the matrix c illustrated in FIG. 43, and uses it for calculating the indirect correction coefficient β. .
When the matrix a indicating that there is no influence between the mailing lists C and D (groups C and D) is obtained by the questionnaire survey, for example, as shown in Expression 32 by the process of S40 described above. From this matrix a, a matrix b is obtained in which the value of an element indicating between these mailing lists C and D (groups C and D) is −1 indicating that there is no influence between them.

In the analysis / evaluation program 50 (FIG. 35), the influence calculation unit 428 of the tabulation / analysis unit 52 (FIG. 36) performs the frequency change determination processing (S26) shown in FIGS. The target scale calculation process (S30) shown in FIGS. 27 to 34 and the process of calculating a numerical value necessary for calculating the degree of influence I using the correction coefficients α and β in the correction processing unit 520 are performed.
The correction processing unit 520 uses the correction coefficients α and β input from the correction coefficient calculation unit 350 to correct the degree of influence I as shown in Equation 28 and Equation 30, and outputs the correction to the result management unit 440. To do.

With reference to FIGS. 45 and 46, the calculation process (S44) of the indirect correction coefficient β by the correction coefficient calculation unit 350 will be further described.
FIG. 45 is a flowchart showing an indirect correction coefficient β calculation process (S44) by the correction coefficient calculation unit 350 shown in FIG.
46 is a chart showing indirect correction coefficient β calculation processing (S44) by the correction coefficient calculation unit 350 shown in FIG.
In addition, 1-7 shown to FIG. 45, FIG. 46 show the correspondence of the process of FIG.

As shown in FIGS. 45 and 46, in step 440 (S440), the correction coefficient calculation unit 350 performs the mailing lists A to D (groups A to D) shown in the left chart of FIG. 46 (same as chart a in FIG. 43). ) Member list, a non-affiliated group list indicated by reference numeral 1 in the chart on the right side of FIG. 46 is created.
In step 442 (S442), the correction coefficient calculation unit 350 aggregates the mailing list (group) in which each member (member) is affected, as indicated by reference numeral 2 in the chart on the right side of FIG.

In step 444 (S444), the correction coefficient calculation unit 350 calculates the correction coefficient β ′ as indicated by the reference numeral 3 in the chart on the right side of FIG.
In step 446 (S446), the correction coefficient calculation unit 350 creates a list of mailing lists (groups) to which each member (member) belongs, as indicated by the reference numeral 4 in the chart on the right side of FIG.
In step 448 (S448), the correction coefficient calculation unit 350, for each member (member), shows a mailing list (group; non-member) to which each member (member) does not belong, as shown in the chart on the right side of FIG. Calculate the number of hops (shortest path) to each group.

In step 448 (S448), the correction coefficient calculation unit 350 adds an average hop to a mailing list (group; non-affiliation group) to which each member (member) does not belong, as indicated by the reference numeral 6 in the chart on the right side of FIG. The number (average shortest path (n + 1)) is calculated.
In step 450 (S450), the correction coefficient calculation unit 350 calculates the indirect correction coefficient β as indicated by the reference numeral 7 in the chart on the right side of FIG.
The process of S446 is a process necessary for obtaining the number of hops (shortest path) from the mailing list (group) to which each member (member) belongs, and can be performed simultaneously with the process of S440.

[Overall operation of network system 1]
The overall operation of the network system 1 (FIG. 1) to which the analysis / evaluation program 50 (FIG. 35) is applied will be described below.
Organizations on the network system 1 (for example, mailing lists A, B, and C assembled in the network system 1) and members of these mailing lists A, B, and C transmit messages by electronic mail.
The analysis / organization apparatus 3 stores a message transmitted on the mailing lists A, B, and C in the log DB 304 as a message log (FIG. 9; organization communication information).

The user of the analysis / organization apparatus 3 activates the analysis / evaluation program 50 (FIGS. 35 and 36), the concept (words, etc.) to be evaluated, the period and organization to be evaluated (mailing list (group)) Is specified, and the process of evaluating the influence of the expression is executed.
As shown in FIG. 39, the questionnaire survey unit 340 of the analysis / evaluation program 50 performs a questionnaire survey on members (members) of the network system 1 and stores the survey results in the questionnaire result DB 342.

The correction coefficient calculation unit 350 calculates the correction coefficients α and β from the questionnaire result stored in the questionnaire result DB 342 and the message log stored in the log DB 304 as shown in FIG.
The aggregation / analysis unit 52 (FIG. 36) reads and analyzes the message log from the log DB 304, and from which of the mailing lists A, B, and C the characteristic expression (determined expression) affects. And the degree of influence of characteristic expressions (deterministic expressions).
Further, the tabulation / analysis unit 52 corrects the obtained degree of influence using the calculated correction coefficients α and β, stores and manages them in the analysis / evaluation DB 308, and corrects them according to user operations. Display or output analysis results.

  The present invention can be used to evaluate information communicated between multiple organizations.

It is a figure which illustrates the structure of the network system with which the evaluation method concerning this invention is applied. FIG. 2 is a diagram illustrating a hardware configuration of a client computer, a server, and an analysis / evaluation apparatus 3 illustrated in FIG. 1. It is a figure which shows the structure of the client program 22 which operate | moves on the client computer shown in FIG. 1, FIG. It is a figure which shows the structure of the server program which operate | moves on the server shown in FIG. 1, FIG. FIG. 5 is a diagram illustrating an e-mail message log recorded in a log DB illustrated in FIG. 4. It is a figure which shows the structure of the 1st analysis and evaluation program which operate | moves on the analysis and evaluation apparatus shown in FIG. 1, FIG. FIG. 7 is a diagram showing organization information stored in an organization / personal DB shown in FIG. 6. FIG. 7 is a diagram showing personal information stored in the organization / personal DB shown in FIG. 6. It is a figure which shows the organization communication information (organization communication information) which a log collection and management part memorize | stores in log DB. It is a figure which shows the structure of the 1st analysis and evaluation part shown in FIG. It is a flowchart which shows the whole process (S10) of the analysis and evaluation part 4. FIG. It is a flowchart which shows the individual total process (S12) shown in FIG. It is a figure which illustrates the feature expression list obtained by totaling expression (noun) by mailing list A (group A) separately. It is a figure which illustrates the feature expression list obtained by totaling an expression element individually by mailing list B (group B). It is a figure which illustrates the feature expression list obtained by totaling an expression element separately with the mailing list C (group C). It is a flowchart which shows the total area determination process (S14) shown in FIG. It is a flowchart which shows the duplication expression extraction process (S16) shown in FIG. It is a figure which illustrates the candidate list in preparation by the process of duplication expression extraction process (S16). It is a figure which illustrates the candidate list | wrist obtained by the process of FIG. 17, FIG. FIG. 17 is a flowchart showing a duplicated expression number determination process (S18) shown in FIG. It is a flowchart which shows the contingency table total process (S22) shown in FIG. It is a figure which illustrates the total result of each mailing list A (group A) obtained by the contingency table total process (S22) shown in FIG. 11, FIG. It is a figure which illustrates the total result of each mailing list B (group B) obtained by the contingency table total process (S22) shown in FIG. 11, FIG. It is a figure which illustrates the total result of each mailing list C (group C) obtained by the contingency table total process (S22) shown in FIG. 11, FIG. FIG. 25 is a flowchart showing the frequency change determination process (S26) shown in FIG. FIG. 26 is a diagram illustrating a frequency change flag table created by the frequency change determination process (S26) shown in FIGS. 11 and 25; It is a flowchart which shows the object scale calculation process (S30) shown in FIG. It is a flowchart which shows the participant list creation process (S32) shown in FIG. It is a figure which illustrates personal ID of each member of mailing list A, B, and C (group A, B, C). It is a figure which illustrates the participant list created by participant list generation processing (S32; FIG. 27, FIG. 28). It is a figure which illustrates the member which overlaps and belongs to mailing list A, B, C (group A, B, C). It is a figure which illustrates the category of the member who belongs to mailing list A, B, C (group A, B, C) redundantly. It is a figure which illustrates the classification table obtained by a category total process (S36; FIG. 27, FIG. 34). It is a flowchart which shows the category total process (S36) shown in FIG. FIG. 7 is a diagram illustrating a configuration of a second analysis / evaluation program used in place of the first analysis / evaluation program illustrated in FIG. 6 or the like in the analysis / evaluation apparatus illustrated in FIG. 1. It is a figure which shows the structure of the 2nd totaling / analysis part of the analysis / evaluation program shown in FIG. It is a figure which illustrates two mailing lists A and B (group A, B) with which an affiliated member (member) has duplication. It is a figure which shows the calculation method of the direct correction variable (alpha) shown to Formula 28. FIG. FIG. 37 is a diagram illustrating a questionnaire GUI (Graphical User Interface) image displayed on the display / input device (FIG. 2) of the client computer (FIG. 1) by the questionnaire survey unit of the analysis / evaluation program shown in FIG. 36. It is a figure which shows three mailing lists A, B, and C (group A, B, and C) which have an overlapping category. It is a figure which shows the indirect correction coefficient (beta) shown in Formula 30 grade | etc.,. FIG. 42 is a diagram illustrating mailing lists A to D (groups A to D) that are affected in a state where a plurality of mailing lists (groups) can intervene. FIG. 43 is a first diagram illustrating a specific example of processing for obtaining an indirect influence degree β from a questionnaire survey in the mailing list (group) illustrated in FIG. 42. It is a flowchart which shows the calculation process (S40) of the hop number by the correction coefficient calculation part shown in FIG. It is a flowchart which shows the calculation process (S44) of the indirect correction coefficient (beta) by the correction coefficient calculation part shown in FIG. 36 is a chart showing indirect correction coefficient β calculation processing (S44) by the correction coefficient calculation unit shown in FIG.

Explanation of symbols

1 ... Network system,
100 ... Network,
2 ... departmental system,
102 ... Departmental LAN,
20: Client computer,
200 ... body,
202... CPU
204: Memory,
206 ... display / input device,
208... Recording device,
210 ... recording medium,
212 ... Communication device,
22 ... Client program,
220 ... UI unit,
222 ... mail program,
224 ... Web browser,
226: LAN communication control unit,
24 ... server,
26: Server program,
260 ... LAN communication control,
262 ... Network communication control,
264 ... Mail server program,
266 ... Web server,
268: Log management unit,
3. Analysis / evaluation equipment,
30, 52 ... Analysis and evaluation program,
300: Network communication control,
302 ... Log collection / management section,
304 ... log DB,
308 ... Analysis / evaluation result DB,
310 ... organization / personal database,
312 ... UI unit,
340 ... Questionnaire survey department,
342 ... Questionnaire result DB,
350 ... correction coefficient calculation unit,
4 ... Analysis / evaluation unit 400 ... Evaluation target range determination unit,
402 ... Member classification part,
420 ... Pre-processing unit,
422 ... Totaling unit,
424 ... Feature extraction unit,
426 ... section determination unit,
428 ... Influence calculation unit,
440 ... Result management unit,
520 ... Correction processing unit,

Claims (7)

An evaluation device realized by a computer having an input / output device, an arithmetic device, and a storage device,
Transmission information storage means for storing transmission information data indicating information transmitted between a plurality of evaluation target units that can be evaluated;
Aggregate the number of appearances of each of the plurality of expressions included in the plurality of stored transmission information data, extract a predetermined number of expressions having the highest number of appearances as feature expressions, all in each period to be evaluated A discriminant that is a function of the difference between the ratio of the number of appearances of each feature expression to the number of appearances of the expression and the probability that each feature expression appears in the entire processing target period including the period to be evaluated. first evaluation target unit of the change in the number of occurrences of each feature representation detects whether arising in period of the evaluation, before Symbol plurality of evaluation target unit for each of the plurality of evaluation units using The feature expression that appears in the first period among the periods in which the change occurs in the first evaluation period in the period in which the change has occurred in the second evaluation object unit of the plurality of evaluation object units Later If it is extracted as a feature representation appearing in two periods, the feature representation is determined to be propagated to the second evaluation object unit from the first evaluation object units, the first evaluation object units wherein An evaluation unit that evaluates that the second evaluation target unit is affected and calculates an evaluation result based on a change in the number of appearances of the feature expression;
An image for investigation for investigating which of the other evaluation target units has influenced each of the constituent elements of the evaluation target unit is displayed on a terminal relating to a plurality of constituent elements included in the plurality of evaluation target units. , using the survey image, by operation of the terminal, influence the evaluation unit given to obtain the survey results produced so as to be selected by the user of the terminal, the findings obtained An investigation means to memorize,
Calculating a first correction coefficient on the basis of the investigation result stored in the investigation means for the first component included in both the first evaluation object unit and the second evaluation object unit; The second correction coefficient is calculated by dividing the number of the evaluation target units corresponding to the transmission destination of the transmission information data transmitted from the first component by the number of the evaluation targets including the first component. Correction coefficient calculation means for
An evaluation apparatus comprising: correction means for correcting the evaluation result based on a third correction coefficient obtained by multiplying the calculated first correction coefficient and second correction coefficient. The correction means includes
Based on the investigation result, based on the number of the evaluation target units interposed in the propagation path of the feature expression between the first evaluation target unit and the second evaluation target unit, An evaluation target unit intervening in the propagation path of the feature expression, and obtaining an average value of the degree of influence between the evaluation target units that directly affect,
From the first evaluation target unit to the second evaluation target unit based on the number of the evaluation units intervening in the propagation path of the obtained characteristic expression and the average value of the obtained influence degree. The evaluation apparatus according to claim 1, wherein the evaluation result relating to the influence is corrected. The correction means is based on a coefficient obtained as a power root of a numerical value corresponding to the number of the evaluation units intervening in the propagation path of the obtained characteristic expression of the average value of the obtained influence degree. The evaluation apparatus according to claim 2, wherein the evaluation result relating to the influence from the evaluation target unit to the second evaluation target unit is corrected. An evaluation method realized by a computer having an input / output device, an arithmetic device, and a storage device,
The transmission information storage means realized by the computer stores transmission information data indicating information transmitted among a plurality of evaluation target units that can be evaluated.
The evaluation means realized by the computer totals the number of appearances of each of the plurality of expressions included in the plurality of stored transmission information data, and extracts a predetermined number of expressions having the highest number of appearances as feature expressions. The ratio of the number of appearances of each feature expression to the number of appearances of all the expressions in each period to be evaluated and the probability that each feature expression will appear in all the processing target periods including the period to be evaluated using a function in which discriminant of said change in the number of occurrences of each feature representation detects whether arising in period of the evaluation for each of the plurality of evaluation units and, before Symbol plurality evaluated the appearance feature representation for a first period of time in which the change has occurred in the first evaluation object unit of the unit in the second evaluation object unit of the plurality of evaluation units Wherein when it is extracted as a feature representation appearing in the second period after the first period, the second evaluation object the feature representation from the first evaluation object unit of the period of the resulting It is determined that the first evaluation target unit has influenced the second evaluation target unit, and an evaluation result is calculated based on a change in the number of appearances of the feature expression,
The plurality of evaluation target units includes an investigation image for investigating whether each of the constituent elements of the evaluation target unit is influenced by which other evaluation target unit, the investigation means realized by the computer A survey result created by causing a user of the terminal to select an evaluation target unit that has been affected by an operation by the terminal using the survey image. Acquire and store the acquired survey results,
The correction coefficient calculation means realized by the computer performs a first operation based on the stored investigation result regarding the first component included in both the first evaluation target unit and the second evaluation target unit. A correction coefficient of 1 is calculated, and the number of the evaluation target units corresponding to the transmission destination of the transmission information data transmitted from the first component is divided by the number of the evaluation targets including the first component. To calculate the second correction coefficient,
An evaluation method in which the correction means realized by the computer corrects the evaluation result based on a third correction coefficient obtained by multiplying the calculated first correction coefficient and second correction coefficient. A program to be executed by a computer having an input / output device, an arithmetic device, and a storage device,
A transmission information storage step for storing transmission information data indicating information transmitted between a plurality of evaluation target units that can be evaluated;
Aggregate the number of appearances of each of the plurality of expressions included in the plurality of stored transmission information data, extract a predetermined number of expressions having the highest number of appearances as feature expressions, all in each period to be evaluated A discriminant that is a function of the difference between the ratio of the number of appearances of each feature expression to the number of appearances of the expression and the probability that each feature expression appears in the entire processing target period including the period to be evaluated. first evaluation target unit of the change in the number of occurrences of each feature representation detects whether arising in period of the evaluation, before Symbol plurality of evaluation target unit for each of the plurality of evaluation units using The feature expression that appears in the first period among the periods in which the change occurs in the first evaluation period in the period in which the change has occurred in the second evaluation object unit of the plurality of evaluation object units Later If it is extracted as a feature representation appearing in two periods, the feature representation is determined to be propagated to the second evaluation object unit from the first evaluation object units, the first evaluation object units wherein An evaluation step for evaluating that the second evaluation target unit is affected and calculating an evaluation result based on a change in the number of appearances of the feature expression;
An image for investigation for investigating which of the other evaluation target units has influenced each of the constituent elements of the evaluation target unit is displayed on a terminal relating to a plurality of constituent elements included in the plurality of evaluation target units. , using the survey image, by operation of the terminal, influence the evaluation unit given to obtain the survey results produced so as to be selected by the user of the terminal, the findings obtained An investigation step to remember,
A first correction coefficient is calculated on the basis of the stored investigation result regarding the first component included in both the first evaluation target unit and the second evaluation target unit, and the first correction coefficient is calculated. A correction coefficient for calculating a second correction coefficient by dividing the number of the evaluation target units corresponding to the transmission destination of the transmission information data transmitted from the component by the number of the evaluation targets including the first component A calculation step;
A program that causes the computer to execute a correction step of correcting the evaluation result based on a third correction coefficient obtained by multiplying the calculated first correction coefficient and second correction coefficient. The correction step includes
Based on the investigation result, based on the number of the evaluation target units interposed in the propagation path of the feature expression between the first evaluation target unit and the second evaluation target unit, An evaluation target unit intervening in the propagation path of the feature expression, and obtaining an average value of the degree of influence between the evaluation target units that directly affect,
From the first evaluation target unit to the second evaluation target unit based on the number of the evaluation units intervening in the propagation path of the obtained characteristic expression and the average value of the obtained influence degree. The program according to claim 5, wherein the evaluation result relating to the influence is corrected. The correction step is based on a coefficient obtained as a power root of a numerical value corresponding to the number of the evaluation units intervening in the propagation path of the obtained characteristic expression of the average value of the obtained influence degree. The program according to claim 6, wherein the evaluation result relating to the influence from the evaluation target unit to the second evaluation target unit is corrected.

Images