JP5444769B2 - Team extraction device - Google Patents

Description

  The present invention relates to a team extraction device, and more particularly to a team extraction device that extracts a team in a company or the like.

  Peter Drucker said, “In a new economy, knowledge is just one more meaningful resource than labor, capital and land as traditional production factors. In knowledge economy, only knowledge is new. It is a source of value ”(for example, see Non-Patent Document 1).

  However, since this knowledge, which is the only management resource, can easily become obsolete, companies are required to constantly create new knowledge within the organization in order to maintain competitiveness. Organizations in are now required to transform themselves into knowledge creation creation organizations.

  The introduction of a knowledge management system can be considered as a measure for an organization to change to a knowledge-creating type, but since the method for measuring and evaluating the effect of the introduced knowledge management has not been established, the cost-effectiveness of the introduction of knowledge management is analyzed. As a result, it has become a major impediment to the introduction of knowledge management by companies.

  In order to establish a method for measuring and evaluating the effects of knowledge management, it is first necessary to grasp and analyze knowledge creation activities in the organization. In particular, communication activities in organizations have attracted attention as knowledge creation activities in recent years, and methods for grasping and analyzing communication activities in organizations have been proposed.

  For example, dialogue behavior information is extracted from sharing physical fields, electronic mail and messenger exchanges, and conducting electronic conferences. By analyzing and comparing the amount of information implemented outside the organization, and analyzing the dialogue behavior in the organization by discriminating the progress of activities across the organization, the organization manager can be informed about the current state of the dialogue behavior in the organization. There is one that allows grasping and promoting dialogue behavior between organizations (for example, see Patent Document 1).

  In addition, by calculating the amount of electronic communication between organizations, the physical distance between organizations, the organizational distance between organizations, and the process distance between organizations as indicators, and extracting and visualizing the correlation between the indicators, Some have the status of electronic communication in the organization grasped and determined utilization guidelines (for example, see Patent Document 2).

  In addition, it is judged whether the electronic mail exchanged in the organization is a new mail, a reply mail, or a forwarding mail, and by grasping the activity state of the electronic mail in the organization based on the usage ratio thereof, the organization Some of them analyze the characteristics of communication (see, for example, Patent Document 3).

  In the conventional technology described above, the organization is grasped in a formal framework such as XX department, XX section or XX project. However, innovative activities in companies are not activities in official organizations, but in “activities performed by members gathered with voluntary intentions that do not have the authority or role prescribed by the official organization (shadow work)” (For example, refer nonpatent literature 2).

  In other words, it is not enough for the organization to analyze the knowledge creation activities in the organization to cover the formal organization, and there are also “hidden” organizations that do shadow work, that is, implicitly active teams. Although it is necessary to make it a target, the above-described conventional technique has a problem that it is not possible to extract an implicitly active team.

  Therefore, an object of the present invention is to provide a team extraction device that can extract a team that operates implicitly.

The team extraction device of the present invention extracts a mail log database in which a mail log representing a mail transmission / reception history is recorded, and a response relation pair representing a mail pair in a response relation from the mail log recorded in the mail log database. A response relationship pair extraction unit, a response relationship chain extraction unit for extracting a response relationship chain representing a chain of mail response relationships based on the response relationship pair, and a sender and a receiver of the mail constituting the response relationship chain A response relationship chain vector generation unit that generates a response relationship chain vector based on the appearance frequency of the response, and a response that represents a set of response relationship chains similar to each other by classifying the response relationship chain based on the response relationship chain vector A response relation chain set generation unit for generating a relation chain set, and a mail constituting the response relation chain included in the response relation chain set And team extractor from followers and recipient extracts the important user in the response relationship chain set as a team, the team information output unit for displaying and outputting the team information representing the user to configure the information and the team identifying the team And a user information storage unit that associates and stores each user name and affiliation information indicating the affiliation of the user, and is extracted by the team extraction unit when the team information is output and displayed by the team information output unit For the other teams, a display indicating that the users constituting the same team are the same team, based on the affiliation information stored in the user information storage unit, between the user names belonging to the same affiliation, It has a configuration for displaying that they belong to the same affiliation .

  With this configuration, the team extracting apparatus of the present invention extracts the response relation chain of mail from the mail log, and extracts and extracts the response relation chain set including similar response relation chains by classifying the extracted response relation chains. Since important users in the response relation chain set are extracted as teams, it is possible to extract an implicitly active team.

  The team extraction unit includes an adjacency matrix generation unit that generates an adjacency matrix representing a transmission / reception relationship between a sender and a receiver of a mail constituting a response relationship chain included in the response relationship chain set; An importance calculating unit that calculates the importance of each user in the response relation chain set from the matrix, and the users constituting the team may be extracted based on the importance of each user.

  With this configuration, the team extracting apparatus of the present invention generates an adjacency matrix that represents a mail transmission / reception relationship between users from the mail constituting the response relation chain included in the response relation chain set, and calculated from the generated adjacency matrix. Since the users constituting the team are extracted based on the importance of the users, the important users in the response relation chain set can be extracted as the team.

  The importance calculation unit may calculate the centrality of each user based on a network centrality measure as the importance.

  With this configuration, the team extraction device of the present invention uses the network centrality scale calculation method as the process of calculating the importance of the user, and therefore calculates the importance with high validity and reliability for each user. Can do.

  The importance calculation unit may calculate the centrality of each user obtained by the PageRank method as the importance.

  With this configuration, the team extraction apparatus of the present invention uses the PageRank method as a process for calculating the importance of the user, and therefore can calculate the importance with high validity and reliability for each user.

In addition, the team extraction program of the present invention includes a recording step of recording a mail log representing a mail transmission / reception history in a mail log database, and a response relationship representing a mail pair in a response relationship from the mail log recorded in the mail log database. A response relationship pair extracting step for extracting a pair, a response relationship chain extracting step for extracting a response relationship chain representing a chain of mail response relationships based on the response relationship pair, and transmission of a mail constituting the response relationship chain A response relationship chain vector generation step for generating a response relationship chain vector based on the appearance frequency of the receiver and the receiver, and by classifying the response relationship chain based on the response relationship chain vector, A response relationship chain set generation step for generating a response relationship chain set representing a set, and the response relationship And team extraction step of the senders and recipients of mail which constitutes a response relationships chain contained in the chain set to extract an important user in the response relationship chain set as a team,
A team information output step for displaying and outputting team information representing information identifying the team and users constituting the team;
A user information storage step of storing each user name and affiliation information indicating the affiliation of the user in association with each other,
When outputting and displaying the team information in the team information output step, with respect to the team extracted by the team extraction step, a display indicating that the users configuring the same team are the same team,
Based on the affiliation information stored in the user information storage step, a display indicating the same affiliation is made between user names having the same affiliation .

  Therefore, the team extraction program of the present invention extracts a response relationship chain of mails from the mail log, extracts a response relationship chain set composed of similar response relationship chains by classifying the extracted response relationship chains, and extracts the extracted response relationships. Since important users in the linked set are extracted as teams, it is possible to extract teams that are implicitly active.

  In addition, the team extraction step includes an adjacency matrix generation step for generating an adjacency matrix representing a transmission / reception relationship between a sender and a receiver of the mail constituting the response relation chain included in the response relation chain set, and the adjacency matrix generation step. An importance level calculating step of calculating the importance level of each user in the response relation chain set from a matrix, and the users constituting the team may be extracted based on the importance level of each user.

  Therefore, the team extraction program of the present invention generates an adjacency matrix representing the mail transmission / reception relationship between users from the mail constituting the response relation chain included in the response relation chain set, and calculates each user's calculated from the generated adjacency matrix. Since the users constituting the team are extracted based on the importance, it is possible to extract important users as teams in the response relation chain set.

  In the importance calculation step, the centrality of each user based on the network centrality measure may be calculated as the importance.

  Therefore, since the team extraction program of the present invention uses the network centrality scale calculation method as the process of calculating the importance of the user, it is possible to calculate the importance having high validity and reliability for each user. .

  In the importance calculation step, the centrality of each user obtained by the PageRank method may be calculated as the importance.

  Therefore, since the team extraction program of the present invention uses the PageRank method as a process of calculating the importance of the user, it is possible to calculate the importance having high validity and reliability for each user.

  The present invention can provide a team extraction device that can extract a team that operates implicitly.

It is a functional block diagram of a team extraction device as one embodiment of the present invention. It is a hardware block diagram of the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the transmission mail log table recorded on the mail log database which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the reception mail log table recorded on the mail log database which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the user information table recorded on the mail log database which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the response relationship pair extraction process of the response relationship pair extraction part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the response relationship pair list obtained by the response relationship pair extraction part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the response relationship chain extraction process of the response relationship chain extraction part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the response relation chain list produced | generated by the response relation chain extraction part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the response relation chain vector generation process of the response relation chain vector generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the response relation chain vector produced | generated by the response relation chain vector production | generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the response relation chain set production | generation process of the response relation chain set production | generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the response relation chain set produced | generated by the response relation chain set production | generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the adjacency matrix produced | generated by the adjacency matrix production | generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the adjacency matrix production | generation process of the adjacency matrix production | generation part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the importance calculation process of the importance calculation part which comprises the team extraction apparatus as one embodiment of this invention. It is a flowchart for demonstrating the team extraction process of the team extraction part which comprises the team extraction apparatus as one embodiment of this invention, and a team information recording part. It is a figure which shows the example of the table showing the importance of each user calculated by the importance calculation part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the table showing the team information recorded on the team information database which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the example of the user information table recorded on the team information database which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the 1st example of a display by the team information output part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the 2nd example of a display by the team information output part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the 3rd example of a display by the team information output part which comprises the team extraction apparatus as one embodiment of this invention. It is a figure which shows the 4th example of a display by the team information output part which comprises the team extraction apparatus as one embodiment of this invention.

  Embodiments of the present invention will be described below with reference to the drawings.

  As shown in FIG. 1, a team extraction apparatus 1 according to an embodiment of the present invention includes a mail log database 10 in which a mail log representing a mail transmission / reception history is recorded, and a mail log recorded in the mail log database 10. A response relationship pair extraction unit 11 for extracting a response relationship pair representing a mail pair in response relationship, a response relationship chain extraction unit 12 for extracting a response relationship chain representing a mail response relationship chain, and a response relationship chain are configured. A response relationship chain vector generation unit 13 for generating a response relationship chain vector based on the appearance frequency of the sender and receiver of the mail to be sent, and a response similar to each other by classifying the response relationship chain based on the response relationship chain vector A response relationship chain set generation unit 14 for generating a response relationship chain set representing a set of relationship chains, and an important user in the response relationship chain set. A team extraction unit 15 for extracting a team, a team information database 16 in which team information representing information for identifying a team and users constituting the team is recorded, and a team information record for recording team information in the team information database 16 A unit 17 and a team information output unit 18 that outputs team information stored in the team information database 16 are provided.

  More specifically, as shown in FIG. 2, the team extraction apparatus 1 includes a plurality of client apparatuses 3a to 3c and a server apparatus 5 that is connected to the client apparatuses 3a to 3c via a corporate network 4 so that they can communicate with each other. Consists of.

  In FIG. 2, three client apparatuses 3a to 3c are shown, but the number is not limited. Further, as shown in FIG. 2, another device such as a printer device 6 may be connected to the corporate network 4.

  Each of the client devices 3a to 3c and the server device 5 is generally provided with a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), a hard disk device, an input device, a display device, a communication module, and the like. It is constituted by a simple computer device.

  In the present embodiment, the mail log database 10 and the team information database 16 are configured by a hard disk device of the server device 5. In addition, the response relationship pair extraction unit 11, the response relationship chain extraction unit 12, the response relationship chain vector generation unit 13, the response relationship chain set generation unit 14, the team extraction unit 15, and the team information recording unit 17 use the server device 5 as each means. It is comprised by CPU which runs the program for functioning as 11 thru | or 15 and 17.

  The team information output unit 18 includes a CPU that executes a program for causing the client devices 3a to 3c or the server device 5 to function as the team information output unit 18, and a display device. The team information output unit 18 may be configured by the printer device 6.

  In the present embodiment, each time the user transmits / receives mail using the mail client software installed in the client apparatuses 3a to 3c, the server apparatus 5 displays the title, sender, receiver, and transmission date / time of the mail. Information to be expressed is acquired, and a mail log with an identifier (hereinafter referred to as “mail ID”) attached to the acquired information is recorded in the mail log database 10. In the present embodiment, since the mail text information is not required, the mail text information is not included in the mail log.

  The server device 5 may be configured by a mail server. In this case, the server device 5 is configured to record the above-described mail log in the mail log database 10 every time the mail transmitted / received by the client devices 3a to 3c is relayed.

  When the server device 5 is accessible to a mail server (not shown), the mail log obtained from the mail server may be recorded in the mail log database 10.

  The mail log database 10 is configured by a general database, and may be configured by an existing database system such as PostgreSQL or MySQL, for example.

  In the mail log database 10, a transmission mail log table as shown in FIG. 3 is recorded. The transmission mail log table represents the mail ID, title, sender user identifier (hereinafter, the user identifier is simply referred to as “user ID”, and the sender user ID is simply referred to as “sender ID”) and the transmission date and time. Consists of lines. It is assumed that the mail ID is assigned in order from 1 in the order of transmission date and time.

  For example, in the transmission mail log table shown in FIG. 3, the first line has a mail ID of 1, a title of “Matter of plan”, a sender ID of 1, and a transmission date and time of May 11, 2008 10:10. Indicates minutes.

  The mail log database 10 stores a received mail log table as shown in FIG. The reception mail log table includes a mail ID, a user ID of the receiver (hereinafter simply referred to as “recipient ID”), and a row indicating a reception type. For example, in the received mail log table shown in FIG. 4, the first line indicates that the mail ID is 1, the receiver ID is 2, and the reception type is to.

  Since there may be a plurality of mail recipients for one mail, there may be a plurality of lines having the same mail ID in the received mail log table. In FIG. 4, the reception mail log table includes the reception type, but this may be omitted.

  In the mail log database 10, a user information table as shown in FIG. 5 is recorded in advance. The user information table includes user IDs, user names, mail addresses, and rows representing affiliations.

  For example, in the user information table shown in FIG. 5, the first line has the user ID 1, the user name “Yamada”, the mail address “yamada@dev.abc.com”, and the affiliation “development section”. It is shown that.

  In FIG. 1, the response relationship pair extraction unit 11 determines whether a pair of mails having different mail IDs is in a response relationship among the mail logs recorded in the mail log database 10, and determines that they are in a response relationship. The mail ID pair of the received mail is added to the response relationship pair list as a response relationship pair.

  Here, the response relationship pair extraction unit 11 determines that there is a response relationship when the mail pair satisfies the reply condition or the transfer condition, and when the mail pair does not satisfy either the reply condition or the transfer condition, It is determined that there is no response relationship.

  The response relationship pair extraction unit 11 determines that the mail log satisfies the reply condition when the mail pair satisfies all of the following conditions 1 to 4, and if any of the conditions 1 to 4 is not satisfied, It is determined that the mail log does not satisfy the reply condition. Here, a mail pair is a mail I and mail J pair.

1. The sender of Mail J is included in the set of recipients of Mail I.
2. The sender of Mail I is included in the set of recipients of Mail J.
3. The transmission date and time of Mail I is before the transmission date and time of Mail J.
4). The title of Mail J matches the title of Mail I with the prefix representing the regular expression “/ re: / i” added.

  The response relationship pair extraction unit 11 determines that the mail log satisfies the transfer condition when the mail pair satisfies all of the following conditions 5 to 8, and if the mail log does not satisfy any of the conditions 5 to 8, It is determined that the mail log does not satisfy the transfer condition. Again, the mail pair is the mail I and mail J pair.

5. The sender of Mail J is included in the set of recipients of Mail I.
6). The product of the set of recipients of Mail J and the set of the sender and receiver of Mail I is the empty set.
7). The transmission date of Mail I is before the transmission date of Mail J.
8). The title of Mail J matches the title of Mail I with a prefix representing the regular expression “/ fw: / i” added.

  For example, in the transmission mail log table shown in FIG. 3 and the reception mail log table shown in FIG. 4, first, the response relationship pair extraction unit 11 performs a mail log with a mail ID of 1 (more specifically, a transmission mail log table and a reception mail log). In the table, the mail ID is 1. The mail log of the other mail ID is interpreted in the same manner.), And the mail log with the mail ID of 2 is acquired.

  A set of recipient IDs of a mail with a mail ID of 1 (hereinafter referred to as “mail 1”, hereinafter also expressed in the same manner for mails of other mail IDs) is {2, 7, 8, 3}. Since the sender ID of Mail 2 is 5, Mail 1 and Mail 2 do not satisfy Condition 1. For this reason, the response relationship pair extraction unit 11 determines that Mail 1 and Mail 2 do not satisfy the reply condition.

  Similarly, since the mail 1 and the mail 2 do not satisfy the condition 5, the response relationship pair extraction unit 11 determines that the mail 1 and the mail 2 do not satisfy the transfer condition. As a result, the response relationship pair extraction unit 11 determines that Mail 1 and Mail 2 are not in a response relationship.

  Next, the response relationship pair extraction unit 11 acquires a mail log with a mail ID of 3, where a set of recipient IDs of the mail 1 is {2, 7, 8, 3}, and the sender of the mail 2 Since ID is 2, Mail 1 and Mail 3 satisfy Condition 1.

  Further, since the set of recipient IDs of Mail 3 is {1, 7, 8} and the sender ID of Mail 1 is 1, Mail 1 and Mail 3 satisfy the condition 2. Also, since the transmission time of Mail 1 is 2008-05-11 10:10:00 and the transmission time of Mail 3 is 2008-05-11 11:11:03, Mail 1 and Mail 3 are Condition 3 is satisfied.

  Further, since the title of Mail 1 is “Matter of Plan” and the title of Mail 3 is “Re: Matter of Plan”, Mail 1 and Mail 3 also satisfy Condition 4. Therefore, the response relationship pair extraction unit 11 determines that the mail 1 and the mail 3 satisfy the reply condition.

  As a result, the response relationship pair extraction unit 11 determines that the mail 1 and the mail 3 are in a response relationship, and sets the mail ID pair (1, 3) of the mail 1 and the mail 3 as a response relationship pair. Add to list.

  Similarly, since the mail 3 and the mail 5 satisfy the transfer condition, the response relationship pair extraction unit 11 determines that the mail 3 and the mail 5 are in a response relationship, and the mail ID pair of the mail 3 and the mail 5 is a pair. (3, 5) is added to the response relationship pair list as a response relationship pair.

  The response relationship pair extraction process of the response relationship pair extraction unit 11 will be described with reference to FIG. In FIG. 6, I, J, and N are integer variables.

  First, the response relationship pair extraction unit 11 substitutes 1 for I, and substitutes the maximum value of the mail ID for N, thereby generating an empty response relationship pair list (step S1). Next, the response relationship pair extraction unit 11 determines whether I is N or less (step S2).

  If it is determined that I is not N or less, the response relationship pair extraction unit 11 ends the response relationship pair extraction process. On the other hand, when it is determined that I is N or less, the response relationship pair extraction unit 11 acquires a mail log with a mail ID of I from the mail log database 10 (step S3).

  Next, the response relationship pair extraction unit 11 substitutes I + 1 for J (step S4), and determines whether J is N or less (step S5). If it is determined that J is not N or less, the response relationship pair extraction unit 11 adds 1 to I (step S6), and returns the response relationship pair extraction process to step S2.

  On the other hand, if it is determined that J is N or less, the response relationship pair extraction unit 11 acquires a mail log with a mail ID J from the mail log database 10 (step S7). Next, the response relationship pair extraction unit 11 determines whether or not the mail with the mail IDs I and J is in a response relationship (step S8).

  When it is determined that the mail with the mail IDs I and J is in a response relationship, the response relationship pair extraction unit 11 adds the mail ID pair with I and J to the response relationship pair list ( Step S9).

  In step S8, when it is determined that the mail IDs I and J are not in a response relationship, or after the mail ID pair of I and J is added to the response relationship pair list, the response relationship pair extraction unit 11 Adds 1 to J (step S10), and returns the response relationship pair extraction process to step S5.

  By executing such a response relationship pair extraction process, the response relationship pair extraction unit 11 can obtain a response relationship pair list as shown in FIG. When the amount of the target mail log becomes enormous, a plurality of response relationship pair extraction units 11 may use a database such as the mail log database 10 to distribute processing.

  In FIG. 1, the response relationship chain extraction unit 12 regards the response relationship chain as a communication sequence that continues with one topic, and extracts a response relationship chain list in which response relationships are consecutive based on the response relationship pair list. Is supposed to generate.

  Specifically, the response relationship chain extraction unit 12 sets the response relationship pair list extracted by the response relationship pair extraction unit 11 as the initial value of the response relationship chain list, and the I-th response relationship chain and J For the i th (I <J) response relationship chain, if the element sequence excluding the first element of the I th response relationship chain matches the element sequence excluding the end element of the J th response relationship chain, that is, the chain relationship Is added to the end of the I-th response relationship chain (hereinafter referred to as “connect”), and a new response relationship chain list (hereinafter “extended response”) is added. The response relation chain is extracted by adding to the "relation chain list".

  Note that the response relationship chain extraction unit 12 repeatedly performs the above-described extraction of the response relationship chain until a new response relationship chain cannot be extracted, that is, until a response relationship chain cannot be added to the extended response relationship chain list. It has become.

  The response relationship chain extraction process of the response relationship chain extraction unit 12 will be described with reference to FIG. In FIG. 8, I, J and N are integer variables.

  First, the response relationship chain extraction unit 12 substitutes the response relationship pair list extracted by the response relationship pair extraction unit 11 as an initial value of the response relationship chain list (step S20). Next, the response relationship chain extraction unit 12 generates an empty extended response relationship chain list (step S21), substitutes 1 for I, and substitutes the number of rows in the response relationship chain list for N (step S22).

  Next, the response relationship chain extraction unit 12 determines whether I is N or less (step S23). If it is determined that I is N or less, the response relationship chain extraction unit 12 substitutes I + 1 for J (step S24), and determines whether J is N or less (step S25). ).

  Here, if it is determined that I is N or less, the response relationship chain extraction unit 12 determines whether the Jth response relationship chain is in a chain relationship with the Ith response relationship chain (step) S26).

  If it is determined that the Jth response relationship chain is in a chain relationship with the Ith response relationship chain, the response relationship chain extraction unit 12 converts the Jth response relationship chain into the Ith response relationship chain. Link and add to the extended response relationship chain list (step S27).

  If it is determined in step S26 that the J-th response relationship chain is not in a chain relationship with the I-th response relationship chain, or the J-th response relationship chain is linked to the I-th response relationship chain, and the extended response relationship chain Then, the response relationship chain extraction unit 12 adds 1 to J (step S28), and returns the response relationship chain extraction process to step S25.

  If it is determined in step S25 that I is not N or less, the response relationship chain extraction unit 12 adds 1 to I (step S29), and returns the response relationship chain extraction process to step S23.

  If it is determined in step S23 that I is not N or less, the response relationship chain extraction unit 12 determines whether or not the extended response relationship chain list is empty (step S30), and the extended response relationship chain. If it is determined that the list is empty, the response relationship chain extraction process is terminated.

  On the other hand, if it is determined that the extended response relationship chain list is not empty, the response relationship chain extraction unit 12 deletes the response relationship chain except those not linked to other response relationship chains from the response relationship chain list. (Step S31). Then, the response relationship chain extraction unit 12 adds the extended response relationship chain list to the response relationship chain list (step S32), and returns the response relationship chain extraction process to step S21.

  Here, a specific example of the response relationship chain extraction process of the response relationship chain extraction unit 12 will be described using the response relationship pair list shown in FIG.

  First, the response relationship chain extraction unit 12 adds (1, 3), (2, 6), (3, 5), (3, 10), (5, 9), (10, 15) to the response relationship chain list. ).

  Next, the response relationship chain extraction unit 12 matches the element sequence (3) excluding the first element of (1, 3) with the element sequence (3) excluding the first element of (3, 5). Since there is a relationship, (1, 3, 5) is added to the extended response relationship chain list.

  Similarly, the response relationship chain extraction unit 12 adds (1, 3, 10), (3, 5, 9), and (3, 10, 15) to the extended response relationship chain list. Here, since other than (2, 6) is connected to other response relationship chains, the response relationship chain extracting unit 12 deletes other than (2, 6) from the response relationship chain list.

  Next, the response relationship chain extraction unit 12 adds the extended response relationship chain list (1, 3, 5), (1, 3, 10), (3, 5, 9), (3, 10, 15) to the response relationship chain list. ) Is added, and this extended response relationship chain list is processed as a response relationship chain list.

  The response relationship chain extraction unit 12 extracts three response relationship chains (2, 6), (1, 3, 5, 9), and (1, 3, 10, 15) by repeating such processing. Then, a response relation chain list as shown in FIG. 9 is generated. In FIG. 9, the chain ID is for identifying each response relation chain, and the response relation chain is shown in a form in which the mail ID is separated by a comma.

  In FIG. 1, the response relationship chain vector generation unit 13 generates a response relationship chain vector based on the appearance frequency of the sender and receiver of the mail constituting the response relationship chain extracted by the response relationship chain extraction unit 12. It is like that.

  The response relationship chain vector generation process of the response relationship chain vector generation unit 13 will be described with reference to FIG.

  First, the response relationship chain vector generation unit 13 acquires the response relationship chain list generated by the response relationship chain extraction unit 12 (step S40), and for each response relationship chain constituting the acquired response relationship chain list, The following processing (steps S41 to S43) is executed.

  The response relationship chain vector generation unit 13 refers to the mail log database 10 and counts the appearance frequencies of the senders and receivers of all mails constituting the response relationship chain (step S41).

  Next, the response relationship chain vector generation unit 13 generates a response relationship chain vector having the counted appearance frequency as an element and the sender / sender of all mails represented by the mail log as feature dimensions (step S42).

  Here, since the element value of the response relationship chain vector is the appearance frequency of the sender / sender of all mails represented by the mail log, it corresponds to the user who is neither the sender nor the receiver of the mail constituting the response relationship chain. The dimension element value is zero.

  Next, the response relationship chain vector generation unit 13 combines the generated response relationship chain vector with the generated response relationship chain vector (step S43). The response relationship chain vector generation unit 13 generates a response relationship chain vector by executing the above processing on all the response relationship chains constituting the response relationship chain list.

  In the response relationship chain vector generation process of the response relationship chain vector generation unit 13 described above, when the appearance frequency of the receiver is counted in step S41, it is greater than 0 and less than or equal to 1 depending on the reception type (see FIG. 4). You may make it count with the weight of.

  Here, a specific example of the response relationship chain vector generation process of the response relationship chain vector generation unit 13 will be described using the response relationship pair list shown in FIG. In FIG. 9, it is assumed that the number of all users is 15, and user IDs 1 to 15 are assigned to each user.

  Mail 2 has a sender user ID of 2 and a receiver user ID of {6, 8, 14}. Mail 6 has a sender user ID of 6 and a receiver user ID of 6. It is assumed that {2, 8, 13, 14}.

  First, the response relationship chain vector generation unit 13 acquires a response relationship chain whose chain ID is 1. Here, since the mails constituting the acquired response relationship chain are the mails 2 and 6, the response relationship chain vector generation unit 13 counts the appearance frequencies of all the senders and receivers of the mails 2 and 6.

  As a result, the appearance frequency of the user with the user ID 2 (hereinafter referred to as “user 2”. Hereinafter, the user with the other user ID is also expressed in the same manner) is 2, the appearance frequency of the user 6 is 2, and the user 8 is. The appearance frequency is 2, the appearance frequency of the user 13 is 1, and the appearance frequency of the user 14 is 2.

  Based on this result, the response relationship chain vector generation unit 13 generates the response relationship chain vector (0, 2, 0, 0, 0, 2, 0, 2, 0, 0, 0, 0, 1, 2, 0). Is generated. The response relationship chain vector generation unit 13 similarly generates response relationship chain vectors for the response relationship chains of chain IDs 2 and 3, and as a result, generates the response relationship chain vector shown in FIG. .

  In FIG. 1, the response relationship chain extraction unit 12 regards the response relationship chain as a communication sequence that continues on one topic, but the response relationship chain set generation unit 14 indicates that several sets of similar response relationship chains are included in several topics. In this example, the response-related chain set is regarded as source information for extracting a team.

  For this reason, the response relationship chain set generation unit 14 generates a response relationship chain set by classifying the response relationship chain based on the response relationship chain vector generated by the response relationship chain vector generation unit 13. .

  The response relationship chain set generation process of the response relationship chain set generation unit 14 will be described with reference to FIG.

  First, the response relationship chain set generation unit 14 acquires the response relationship chain vector generated by the response relationship chain vector generation unit 13 (step S50). Next, the response relationship chain set generation unit 14 generates a response relationship chain set by clustering the response relationship chain using the response relationship chain vector (step S51).

  Here, in the present embodiment, the response relation chain set generation unit 14 uses, for example, a known k-mean method (for example, see Non-Patent Document 3) as clustering. The k-mean method is a parametric method, and it is necessary to specify the number of clusters before execution. Therefore, the response relation chain set generation unit 14 may specify a predetermined value as the number of clusters. Alternatively, the number of clusters may be specified using a known optimal estimation method.

  By executing such a response relationship chain set generation process, the response relationship chain set generation unit 14 generates a response relationship chain set as shown in FIG. In FIG. 13, the set ID refers to an identifier for identifying a cluster.

  In the response relationship chain set shown in FIG. 13, the response relationship chain with the chain ID 1 is included in the response relationship chain set with the set ID 1, and the response relationship chain with the chain ID 2 and 3 has the set ID 2 It is included in the response relation chain set.

  In FIG. 1, the team extraction unit 15 is an adjacency representing a transmission / reception relationship between a sender and a receiver of a mail constituting a response-related chain included in the response-related chain set generated by the response-related chain set generating unit 14. An adjacency matrix generation unit 19 that generates a matrix and an importance calculation unit 20 that calculates the importance of each user in the response relation chain set from the adjacency matrix.

  As shown in FIG. 14, the adjacency matrix generation unit 19 generates an adjacency matrix having a number of rows and columns equal to the number of mail senders and receivers constituting the response-related chain included in the response relation chain set. The frequency at which each sender sends a mail to each recipient is assigned to each element of the adjacency matrix.

  The adjacency matrix generation processing of the adjacency matrix generation unit 19 will be described with reference to FIG. The adjacency matrix generation process is executed for each set ID in the response relation chain set.

  First, the adjacency matrix generation unit 19 acquires a target response relationship chain set from the response relationship chain sets generated by the response relationship chain set generation unit 14 (step S60). Next, the adjacency matrix generation unit 19 extracts all mails constituting the response relationship chain included in the response relationship chain set (step S61).

  Next, the adjacency matrix generation unit 19 refers to the mail log database 10 to extract the sender / receiver of the extracted mail (step S62), and has a number of rows and columns equal to the number of the extracted sender / receiver, A square matrix in which all elements are 0 is generated as an adjacency matrix (step S63).

  Next, the adjacency matrix generation unit 19 adds 1 to the elements of the adjacency matrix corresponding to the sender and the receiver for each mail (step S64).

  Here, a specific example of the adjacency matrix generation process of the adjacency matrix generation unit 19 will be described using the response relation chain set shown in FIG. In FIG. 13, it is assumed that the number of all users is 15, and user IDs 1 to 15 are assigned to each user.

  Mail 2 has a sender user ID of 2 and a receiver user ID of {6, 8, 14}. Mail 6 has a sender user ID of 6 and a receiver user ID of 6. It is assumed that {2, 8, 13, 14}.

  First, since the response-related chains included in the response-related chain set with the set ID 1 are only those with the chain ID 1, the adjacency matrix generation unit 19 uses the response-related chain with the chain ID 1 from the response-related chain list. Are extracted.

  Next, the user ID of the sender of Mail 2 is 2, the user ID of the receiver is {6, 8, 14}, the user ID of the sender of Mail 6 is 6, and the user ID of the receiver is {2, 8, 13 and 14}, the ID of the sender / receiver related to the response-related chain set with the set ID 1 is {2, 6, 8, 13, 14}, so the adjacency matrix generation unit 19 has a size of 5 A square matrix of × 5, each row and each column corresponding to {2, 6, 8, 13, 14} and all elements being 0 is generated as an adjacency matrix.

  Here, since the user ID of the sender of Mail 2 is 2 and the user ID of the receiver is {6, 8, 14}, the adjacency matrix generation unit 19 uses (1, 2,), (1 , 3), 1 is added to each element of (1, 5).

  In addition, since the user ID of the sender of the mail 6 is 6 and the user ID of the receiver is {2, 8, 13, 14}, the adjacency matrix generation unit 19 uses the adjacency matrix (2, 1), ( 2, 3), (2, 4), and (2, 5) are each incremented by 1. As a result, the adjacency matrix as shown in FIG. 14 is generated by the adjacency matrix generation unit 19.

  In FIG. 1, the importance calculation unit 20 may calculate the centrality of each user based on a well-known network centrality measure (for example, see Non-Patent Document 4) as the importance, and the well-known PageRank method You may make it calculate the centrality of each user obtained by (for example, refer patent document 4) as importance.

  Such importance calculation processing of the importance calculation unit 20 will be described with reference to FIG.

  First, the importance calculation unit 20 acquires the adjacency matrix generated by the adjacency matrix generation unit 19 (step S70). Next, the importance calculation unit 20 calculates the importance of each user in the response relationship chain set from the adjacency matrix (step S71).

  Here, the team extraction processing of the team extraction unit 15 and the team information recording unit 17 will be described with reference to FIG. The team extraction process is executed for each set ID in the response relation chain set.

  First, the team extraction unit 15 causes the adjacency matrix generation unit 19 to execute adjacency matrix generation processing (see FIG. 15) (step S80), and causes the importance level calculation unit 20 to execute importance level calculation processing (see FIG. 16) (see FIG. 16). Step S81).

  Next, the team extraction unit 15 compares the importance level of each user calculated by the importance level calculation unit 20 with a predetermined threshold value to extract important users in the response relation chain set as a team. (Step S82).

  When a team is extracted by the team extraction unit 15, the team information recording unit 17 records information on the extracted team in the team information database 16 (step S83).

  In the present embodiment, a predetermined threshold is used when an important user in each team is extracted based on the importance of each user calculated by the importance calculation unit 20.

  Here, many well-known methods for calculating the threshold value of importance by statistical processing have been proposed, but these are based on statistical criteria, so that problems occur when applied to actual data. There are many.

  Therefore, in the present embodiment, as a threshold value to be compared with the importance level of each user, for example, “usually a team is composed of 5 or less” or “experience that a page rank value of 0.5 or more is considered important. Use a value determined by empirical knowledge, such as

  FIG. 18 shows an example of a table representing the importance level of each user calculated by the importance level calculation unit 20. Here, the importance in FIG. 18 is a page rank value calculated by the PageRank method, and the threshold is assumed to be 0.5 based on heuristic knowledge.

  First, when a user having an importance of 0.5 or higher is extracted by the team extraction unit 15 in the response-related chain set with the set ID 1, a team composed of users with user IDs 1, 3, and 6 is extracted. Is done.

  Similarly, when a user having an importance level of 0.5 or higher is extracted by the team extraction unit 15 in the response-related chain set with the set ID 2, the user ID is configured with users 1, 2, 4, and 7 Teams are extracted.

  Note that a user with a user ID of 5 does not belong to any team because it has a low importance in the response relationship chained set with the set IDs of 1 and 2, while a user with a user ID of 1 Belongs to both teams because it has high importance in both of the response relation chain sets of 1 and 2.

  FIG. 19 shows an example of a team information table recorded in the team information database 16.

  This table includes a team ID representing a team identifier, a user ID, and an importance field. For example, in the first line, a team with a team ID of 1 belongs to a user with a user ID of 1, and users in this team The importance of 1 is 0.65.

  In FIG. 1, the team information output unit 18 outputs the team information stored in the team information database 16. Here, display examples of team information based on the team information table shown in FIG. 19 and the user information table shown in FIG. 20 included in the team information database 16 are shown in FIGS.

  In FIG. 21, the team extracted by the team extraction unit 15 and the user names belonging to the team are displayed. For example, “Yamada”, “Suzuki”, and “Sasaki” belong to the team 1. I understand. Also, by placing each user name at a position where the distance from the center of the circle indicating the team and the importance are inversely proportional, the importance of each user in the team can be indicated. In FIG. 21, user names that do not belong to any team are also displayed.

  Therefore, the team information output unit 18 can cause the administrator to detect an active team that does not match the official affiliation by outputting the team information as shown in FIG. 21, and the importance of the user in the team (Activity level) can be grasped by the administrator.

  In FIG. 22, in addition to the display shown in FIG. 21, the degree of interaction between users is also displayed as the thickness of a line connecting user names. Note that the degree of exchange between users is obtained from the amount of mail sent and received between users.

  For example, FIG. 22 shows that the exchange degree between “Yamada” and “Sasaki” in Team 1 is high, and the exchange degree between “Suzuki” and “Sasaki” is relatively low. Therefore, the team information output unit 18 can cause the administrator to understand the relationship of members in the active team by outputting the team information as shown in FIG.

  In FIG. 23, the user names of overlapping users are displayed so as to overlap with the display shown in FIG. As a result, the degree of overlap of the user's team can be understood. For example, FIG. 23 shows that “Yamada” belongs to both team 1 and team 2.

  Therefore, the team information output unit 18 outputs team information as shown in FIG. 23, thereby allowing the administrator to understand the degree of duplication between teams, thereby detecting an administrator who operates across the organization. it can.

  In FIG. 24, in addition to the display shown in FIG. 23, user names of users who are placed in official affiliations are displayed connected by lines. For example, “Kato”, “Suzuki”, and “Mori” are connected by a line, so that it is understood that they are arranged in the same affiliation.

  Therefore, the team information output unit 18 can output the team information as shown in FIG. 24 to allow the administrator to understand the difference in structure between the actively active team and the official affiliation. Give managers insights into change.

  In the present embodiment, an example in which information is reflected in display characteristics such as color, shape, and size is not performed, but the team information output unit 18 performs display that actively uses these display characteristics. By performing the above, more effective display can be realized.

  As described above, the team information output unit 18 can effectively display the team information extracted by the team extraction unit 15 to allow the administrator to grasp and analyze the (communication) activity and state of the organization.

  As described above, the team extracting apparatus 1 according to an embodiment of the present invention extracts a response relation chain of mail from a mail log, and classifies the extracted response relation chain to make a response composed of similar response relation chains. Since the relation chain set is extracted and important users in the extracted response relation chain set are extracted as teams, it is possible to extract the teams that are implicitly active.

DESCRIPTION OF SYMBOLS 1 Team extraction apparatus 3a, 3b, 3c Client apparatus 4 In-company network 5 Server apparatus 6 Printer apparatus 10 Mail log database 11 Response relation pair extraction part 12 Response relation chain extraction part 13 Response relation chain vector generation part 14 Response relation chain set generation Section 15 Team extraction section 16 Team information database 17 Team information recording section 18 Team information output section 19 Adjacency matrix generation section 20 Importance calculation section

JP 2006-260341 A JP 2006-252222 A JP 2006-252220 A US Pat. No. 6,285,999

P. F. Drucker, “Post Capitalist Society”, Diamond, 1993 Kazuo Ichijo and Shinichiro Tokuoka, “Shadowwork: Organizational strategy to promote knowledge creation”, Toyo Keizai Inc., February 2007 Toshosho Kawaguchi, “Introduction to Multivariate Analysis II”, Morikita Publishing Masahiro Amagasaki, Yutaka Matsuo, Takuichi Nishimura, Hideaki Takeda, "Network generation model with multiple centrality measures", Journal of Intelligent Information Fuzzy Society, 2008

Claims (9)

A mail log database in which a mail log representing a history of mail transmission and reception is recorded;
A response relationship pair extraction unit that extracts a response relationship pair representing a mail pair in a response relationship from the mail log recorded in the mail log database;
A response relationship chain extraction unit that extracts a response relationship chain representing a chain of mail response relationships based on the response relationship pair;
A response relationship chain vector generation unit for generating a response relationship chain vector based on the appearance frequency of the sender and receiver of the mail constituting the response relationship chain;
A response relationship chain set generation unit that generates a response relationship chain set representing a set of response relationship chains that are similar to each other by classifying the response relationship chain based on the response relationship chain vector;
A team extraction unit for extracting important users in the response relationship chain set as a team from the sender and receiver of the mail constituting the response relationship chain included in the response relationship chain set;
A team information output unit for displaying and outputting team information representing information for identifying the team and users constituting the team;
A user information storage unit that associates and stores each user name and affiliation information indicating the affiliation of the user,
When outputting and displaying the team information by the team information output unit, for the team extracted by the team extraction unit, a display indicating that the users configuring the same team are the same team,
A team extraction device that displays the same affiliation between user names having the same affiliation based on the affiliation information stored in the user information storage unit . The team extraction unit
An adjacency matrix generating unit that generates an adjacency matrix representing a transmission / reception relationship between a sender and a receiver of mail constituting a response relation chain included in the response relation chain set;
An importance calculation unit for calculating the importance of each user in the response relation chain set from the adjacency matrix;
The team extracting apparatus according to claim 1, further comprising: extracting users constituting the team based on the importance of each user.   The team extraction device according to claim 2, wherein the importance calculation unit calculates the centrality of each user based on a network centrality measure as the importance.   The team extraction device according to claim 2, wherein the importance calculation unit calculates the centrality of each user obtained by the PageRank method as the importance. A recording step for recording a mail log representing a history of mail transmission / reception in a mail log database;
A response relationship pair extraction step of extracting a response relationship pair representing a mail pair in a response relationship from the mail log recorded in the mail log database;
A response relationship chain extraction step for extracting a response relationship chain representing a chain of mail response relationships based on the response relationship pair;
A response relationship chain vector generation step for generating a response relationship chain vector based on the appearance frequency of the sender and the receiver of the mail constituting the response relationship chain;
A response relationship chain set generation step for generating a response relationship chain set representing a set of response relationship chains similar to each other by classifying the response relationship chain based on the response relationship chain vector;
A team extraction step of extracting important users in the response relationship chain set as a team from the sender and receiver of the mail constituting the response relationship chain included in the response relationship chain set;
A team information output step for displaying and outputting team information representing information identifying the team and users constituting the team;
A program for causing a computer to execute a user information storage step of storing each user name and affiliation information indicating the affiliation of the user in association with each other,
When outputting and displaying the team information in the team information output step, with respect to the team extracted by the team extraction step, a display indicating that the users configuring the same team are the same team,
A team extraction program for displaying the same affiliation between user names having the same affiliation based on the affiliation information stored in the user information storing step . The team extraction step includes
An adjacency matrix generating step for generating an adjacency matrix representing a transmission / reception relationship between a sender and a receiver of mail constituting a response relationship chain included in the response relationship chain set;
Importance calculation step of calculating the importance of each user in the response relation chain set from the adjacency matrix,
The team extraction program according to claim 5, further comprising: extracting users constituting the team based on the importance of each user.   The team extraction program according to claim 6, wherein in the importance calculation step, the centrality of each user based on a network centrality measure is calculated as the importance.   The team extraction program according to claim 6, wherein in the importance calculation step, the centrality of each user obtained by the PageRank method is calculated as the importance.   A recording medium on which the team extraction program according to claim 5 is recorded so as to be readable by a computer.