JP2005339118A - Reputation information calculation device, program, and recording medium - Google Patents

Abstract

Reputation information with high reliability is obtained by dealing with collusion in a novel manner.
An evaluation information input means for recording evaluation information received from an evaluator's terminal in an evaluation information recording means and a presence of evaluation information for an evaluated person by an evaluator with a community participant as a node. Evaluation route searching means 12 for searching for an evaluation route starting from the evaluated person and ending at the evaluator based on the directed graph connected at the edge by the evaluation information recording means 19, and collusion of evaluation information according to the search result of the evaluation route Reputation information calculation means 13 for calculating reputation information based on the collusion degree and evaluation information, and reputation information output for outputting reputation information from reputation information recording means 20 in which reputation information is recorded And means 18.
[Selection] Figure 1

Description

  The present invention relates to an apparatus, a program, and a recording medium for evaluating community participants with each other and calculating a reputation of the participants in a community that conducts business transactions and information exchange on the Internet. Prevent participants from raising the reputation of the participants in an unfair manner by making a good evaluation against each other.

  In recent years, an always-on connection environment to the Internet has become widespread, and for this reason, participants in the Internet have formed a community, and commercial transactions and information exchange on the Internet have rapidly spread. Here, for example, a transaction system represented by Yahoo! Auctions or eBay (www.ebay.com) is used as a commercial transaction on the Internet, and is simply illustrated as shown in FIG. That is, as shown in FIG. 14, in a community on the Internet 1 made up of participants A to E, for example, buying information BI and selling information SI are brought to the market 2, and the buying procedure BP that is paid by contract establishment And selling procedure SP which is shipping. In FIG. 14, a selling / buying transaction is performed with the participant B as the buyer and the participant D as the seller.

  In this case, business transactions on the Internet are generally different from face-to-face transactions, and since the other party is not visible, various frauds are likely to occur, such as the buyer escaping with the price of the product or the seller neglecting to deliver the product. . In order to reduce such fraud, the reputation system (Reputation System) mechanism that inputs evaluation information such as “good” and “bad” for each other's actions after the transaction ends, and aggregates and discloses this information to the user Provided. In FIG. 14, the evaluation of the evaluator D is sent from the evaluator B to the reputation system, with the participant B as the evaluator and the participant D as the evaluator.

  However, in such a simple reputation system, in reality, one participant creates accounts for a number of fictitious participants and conducts a fictitious transaction with a specific participant to create an intangible “good” By entering an assessment, a fictitious collusion attack is possible that illegally increases the reputation of one participant. Further, there is also an attack in which, even if a fictitious account is not created, some participants actually collide and input a “good” evaluation with each other, thereby illegally increasing the evaluation of the collocated participants.

Therefore, in order to reduce such fraud by collusion, in Non-Patent Document 1, clustering analysis is performed based on the evaluation of participants to group people who perform the same evaluation, and those who belong to the same cluster By weighting the evaluation, a method has been proposed in which the evaluation in the case of collusion belongs to another cluster when viewed from the legitimate participant, and as a result, the influence of the reputation value due to collusion is reduced.
Immunizing Online Reputation Reporting Systems Against Unfair Ratings and Discriminatory Behavior, ACM Conference on Electronic Commerce, 2000: Dellarocas; Sloan School Management

 However, in the system shown in Non-Patent Document 1 as described above, since there are few inputs of evaluation information by oneself for the initial user, it is not assigned to an appropriate cluster, and it is good for one of the groups that happen to be collocated. If the evaluation is input, the user may be clustered into the group.

  In addition, the reputation score value based on clustering by the group is an evaluation value viewed from a specific participant, and is not a global evaluation value for the entire community. For example, the membership fee is discounted based on the reputation score value. And the proposed method is difficult to use for the purpose of distributing incentives.

  In other words, even if the system described in Non-Patent Document 1 is adopted in order to deal with fraud by collusion, it has an insufficient clustering and non-global evaluation score, and it is difficult to obtain a highly reliable reputation score. As a result, there is a problem that the above-mentioned collusion is not sufficiently dealt with.

  The present invention has been invented to solve the above-described problems, and it is possible to provide a reputation information calculation device, a program, and a recording medium for obtaining reputation information with high reliability by dealing with collusion by a novel method. For the purpose of provision.

  The present invention was obtained by an evaluator for an evaluator who is a participant in a community formed by a network and who is an evaluator who is a participant in the community to be evaluated. In the reputation information calculation device for calculating evaluation information of the evaluated person by calculating evaluation information of the evaluated person, evaluation information input means for recording the evaluation information received from the terminal of the evaluator in the evaluation information recording means, Based on a directed graph that connects the presence of evaluation information for the evaluated person by the evaluator with a directed edge using the community participants as nodes, the evaluation information includes an evaluation path starting from the evaluated person and ending at the evaluator. The evaluation route search means for searching by the recording means, the collusion degree of the evaluation information is calculated according to the search result of the evaluation route, and the collusion degree and the evaluation information are And reputation information calculation means for calculating Hazuki the reputation information of the Evaluator, the reputation information is characterized in that it comprises, and reputation information output means for outputting the reputation information from the reputation information recording means recorded.

  Create a directed graph with the participants in the Internet community as nodes and the evaluation of the participants as directed edges, and the evaluation path length and the number of evaluation paths from the evaluated person to the evaluator of this directed graph. From the relationship with the person to be evaluated, it is possible to obtain a credibility with higher credibility than before, and as a result, an evaluation value that can be objectively viewed and is resistant to collusion is obtained. be able to.

Preferred embodiments of the present invention will be described below with reference to the drawings.
1 to 13 show an embodiment of the present invention. FIG. 1 is a computer system placed in the market 2 of FIG. 14, for example, and shows the overall configuration of the reputation information calculation apparatus of this embodiment. In the computer system shown in FIG. 1, the input means is evaluation information input means 10, the output means is reputation information output means 18, the control device 11 is evaluation route search means 12 and reputation information calculation means 13, and the storage means is evaluation information recording. The configuration of means 19 and reputation information recording means 20 is shown.

  In this case, as input of this reputation information calculation device, there is a request from the participant or the evaluator described above, but in the reputation information calculation device, when this request is a request to output (acquire) reputation information, Reputation information output processing will be performed, and if the request is evaluation information input by the evaluator, if it is “good” evaluation information, the degree of collusion is calculated and reputation information is calculated, and “bad” In the case of evaluation information, since it is not necessary to calculate the collusion degree, the processing is performed even if there is an input.

  Part of the processing for this request will be described with reference to the flowchart shown in FIG. The reputation information calculation apparatus shown in FIG. 1 accepts a request transmitted from a participant's user terminal using a communication means such as the Internet (step 21). When the received request is an evaluation information input by the evaluator to the evaluated person (step 22), the evaluation information is recorded in the evaluation information recording means 19 by the evaluation information input process (step 24) by the evaluation information input means 10. . Next, an evaluation route search process (step 25) is executed by the evaluation route search means 12, and an evaluation route that is a search result is acquired using the evaluation information stored in the evaluation information recording means 19. Thereafter, the reputation information calculation means 13 calculates reputation information by executing reputation information calculation processing (step 26) using the obtained evaluation route, and the calculated reputation value is stored in the reputation information recording means 20. Store.

  On the other hand, if the received request is a request for acquiring reputation information for a participant (evaluated person), the reputation information output means 18 uses the reputation information of the user ID (evaluated person ID) specified by the request as the reputation information. Search from the recording means 20 and output (step 23). Then, the reputation information is transmitted to the user terminal of the participant who wants to know the reputation information using communication means such as the Internet.

  Returning to FIG. 1, the details of the block configuration will be described. In FIG. 1, an evaluation information input means 10 receives evaluation information for a community participant (evaluated person) input from an evaluator's user terminal via a communication means such as the Internet, and an evaluation information recording means. 19 is recorded. In this case, the evaluation information is composed of a set of <evaluator ID, evaluated person ID, evaluation value>. Here, the evaluator ID is the user ID of the participant who has input the evaluation information, and the evaluated person ID is the user ID of the participant who has been evaluated. The evaluation value is an evaluation by the evaluator ID with respect to the evaluator ID. For example, the evaluation value takes a value in the range of greater than 0 and 1 or less. In this example, the evaluator is determined to be “good” for simplicity. 1 is given to the evaluation value in the case of the evaluation, and 0 is given to the evaluation value in the case of being judged as “bad”.

  The evaluation information input means 10 has a function of preventing a certain participant from impersonating another participant (evaluator) and inputting evaluation information by authenticating the user ID and password of the participant. For this purpose, it is indispensable for the community participants to have means for giving a user ID and a password by means such as member registration in advance. In this case, many methods have already been proposed for user authentication, and various authentication methods can be used.

  When the evaluation information is input by the evaluator, it is conceivable that the evaluation information is input each time the transaction is made with the same trading partner a plurality of times. Considering such a case, the same evaluator can evaluate the same evaluated person a plurality of times. Here, when the same evaluator evaluates the same evaluator a plurality of times, for example, when the number of inputs of “good” evaluation information is n times, the evaluation value is set to 1− (2 ^ -N) [in this formula, the symbol ^ represents an exponent and this formula represents 2 to the power of -n]. Specifically, when “good” is input once, 0.5 when it is input twice (1-1 / 2), 0.75 when it is input twice (1-1 / 4), when it is input three times When it is done (1-1 / 8), it becomes 0.875. The evaluation value itself increases as the number of evaluations by the same evaluator for the same evaluator increases, but the maximum value is a value smaller than 1. In the above example, when the number of evaluations of the same evaluated person by the same evaluator increases, the evaluation value is increased by the above formula (1- (2 ^ -n)). Processing may be performed by limiting the evaluation information of the same evaluated person by the evaluator so that it can be input only once.

  FIG. 3 illustrates the evaluation information in consideration of the number of evaluations, and illustrates the evaluator ID, the evaluated person ID, the number of evaluations, and the evaluation value. By inputting this evaluation information from the evaluation information input means 10, the evaluation information recording means 19 records this evaluation information as an evaluation information record.

  FIG. 5 is a flowchart showing the operation of the evaluation information input means 10. In FIG. 5, first, evaluation information is received from a user terminal of a participant (evaluator) (step 51). The evaluator ID and the evaluated person ID are taken out from the received evaluation information (step 52). The evaluation information record recorded in the evaluation information recording means 19 is searched using the acquired evaluator ID and the evaluated person ID as keys. If an evaluation information record already exists (step 53), the number of evaluations is updated (step 54). In the example of FIG. 3, the above formula (1- (2 ^ -n)) is used. Recalculate the evaluation value and update the evaluation value. If there is no corresponding evaluation information record in the evaluation information recording means 19, a new evaluation information record comprising a set of <evaluator ID, evaluated person ID, number of evaluations, evaluation value> is additionally set (step 55). As described above, nothing is done when “bad” evaluation information is input from the evaluation information input means 10.

Next, the evaluation route searching means 12 in the control device 11 in FIG. 1 will be described. The evaluation route search means 12 is a directed side in which the participants including the evaluator and the evaluated person are set as vertices (nodes), and the evaluation information (evaluation value) of the evaluated person by the evaluator is added to the direction. When the start point corresponding to the evaluator and the end point corresponding to the evaluator are specified, the evaluation path number P ^L at the evaluation path length L (the number of directed sides of the path) from the start point to the end point is specified. Is a means for searching. In this embodiment, every time evaluation information is input from the evaluation information input means 10, an evaluation path search is performed by the evaluation path length acquisition means and the evaluation path number acquisition means 14 with the evaluated person as the starting point and the evaluator as the end point. To search for a so-called reverse path (back link). In this evaluation route search, a list of the number of routes from the length 1 (number of directed sides 1) to the maximum route length (maximum number of directed sides) LMAX specified in advance by the system is acquired. The evaluation path search here is performed by using a directed graph. For this search, many algorithms have already been proposed in graph theory, and these can be used.

  Hereinafter, the evaluation route search by the evaluation route search means 12 will be described in detail with reference to the flowchart of FIG. When the evaluation information is recorded in the evaluation information recording means 19 by the evaluation information input means 10, it is known whether or not there is an evaluation between the specified evaluator and the evaluated person. In this state, as shown in FIG. 6, the evaluation route searching means 12 generates a matrix based on whether or not there is an evaluation between the evaluator and the evaluated person (step 61). That is, as specifically shown in FIG. 7, when the community participants are A to E according to the example of FIG. 14, FIG. 7A shows the evaluation of the evaluators D and E by the evaluator A. The evaluator B evaluates the evaluated persons C, A, and E, the evaluator C evaluates the evaluated persons A and D, the evaluator D evaluates the evaluated person E, and the evaluator. A directed graph is shown in the situation where E is not evaluating. In this directed graph, A to E are nodes, and an arrow when the evaluator in the case of the evaluation “present” is the start point and the evaluated person is the end point is the directed side. However, here, there is no directed edge in the case of “bad” evaluation. Here, if a matrix with the evaluator as a row and the evaluator as a column is generated and the directed edge is “present” as an element, 1 (including the direction), and 0 when there is no directed edge, The matrix shown in FIG. 7B can be formed. Specifically, the generation of the matrix in step 61 is to create the matrix of FIG. 7B. In FIG. 7B, the diagonal elements of the matrix do not have a directed side and are always 0. This is because the evaluator and the evaluator are not the same person.

  Next, in step 62, initial setting is performed, the initial value of the path length is set to 1, and the initial value of the evaluation path number list indicating the number of evaluation paths for each path length is set to [] (empty list).

Thereafter, in step 63, calculation is performed by the evaluation path length acquisition unit and the evaluation path number acquisition unit 14 shown in FIG. 1, and each participation is performed for each path length (= directed path) starting from 1 following the directed side. Calculate the number of directed paths between the parties. Specifically, the operations shown in FIGS. 8 to 10 are performed using the matrix in FIG. 7B. Based on the directed graph shown in FIG. 7A, the number of directed routes with a route length of 1 obtains 8 directed routes based on the matrix of FIG. 7B. This is represented by a matrix Π ⁽¹⁾ in FIG. 8, and a matrix obtained by squaring this matrix is Π ⁽²⁾ in FIG. 9. Here, based on the graph theory of the directed graph, in the relational diagram showing the matrix in FIG. 8, the first power ⁽¹⁾ in Π (1) represents the directed path length consisting of one directed edge of the directed graph. “1” of each element represents “present” of a directed side as shown in FIG. 7B and at the same time represents the number of directed paths. Similarly, in FIG. 9, the squared ⁽²⁾ in Π ⁽²⁾ represents the directed path length composed of two directed edges of the directed graph, and “1” of each element represents one directed path. "2" represents the number of two directed paths. That is, for example, in FIG. 9, “1” in the first row and the fifth column has one directional route number consisting of two directed sides between AEs in accordance with FIG. 7 (A → D in FIG. 7A). → (E path) “2” in 2 rows and 4 columns indicates that there are two effective paths consisting of two directed edges between the BDs in FIG. 7 (B in FIG. 7A). C → D and B → A → D).

FIG. 10 is an explanatory diagram showing a part of calculation when the matrix of FIG. 8 is squared. In FIG. 10, π represents an element of the matrix Π ^(ν) , and π ⁽²⁾ represents FIG. Means the element with the directed path length consisting of the same two directed edges as in ^(2), and π _AB means the element between AB that is A rows and B columns of the matrix . Therefore, the uppermost expression in FIG. 10 is an element of a matrix Π ⁽²⁾ representing a directed path length composed of two directed edges obtained by squaring Π ⁽¹⁾ . This is the number of directed paths. Then, the calculation of the uppermost expression in FIG. 10 is performed with a matrix (element surrounded by a circle in FIG. 8) of 1 row (A row) excluding 2 columns (B column) in FIG. ) Is a product of two columns (B columns) matrix (elements surrounded by a rectangle in FIG. 8). In general terms, the element π _ij ^(ν) representing the number of effective routes of the matrix Π ^(ν) with the effective route length ν is, for example, the i-line as the starting point of the directed route and the end point of the directed route. a vector that can remove the j-th column element from the vector produced by the i-th row of [pi matrix was column j ^(ν-1) and the j-th row of elements from column j of the matrix Π ^(ν-1) What is necessary is just to take the product of the vector which can be deleted.

  Thus, by calculating the formulas in FIG. 10 in order, each element shown in FIG. 9 is calculated in order from the first row. The calculation formula of FIG. 10 describes seven formulas surrounded by a quadrangle shown in FIG. 9, but in the example of FIG. 10, a total of 20 formulas are excluded except for the diagonal element 0 surrounded by a circle. Is calculated. Note that “1” in the first row and the fifth column in FIG. 9 (“1” in the fourth expression in FIG. 10) has one directional route number of directional route lengths consisting of two directed edges (A → D → E), “1” in 2 rows and 1 column (“1” in the fifth expression in FIG. 10) has one number of directed paths (B → C → A), “2” in 2 rows and 4 columns (“2” in the seventh expression in FIG. 10) indicates that there are two directed paths (B → C → D and B → A → D).

FIG. 11 shows an effective graph when the effective path length is increased. FIG. 11A is a directed graph corresponding to FIG. 7A and is created when certain evaluation information is input. Is a directed graph. As described above, this directed graph is a matrix of Π ⁽¹⁾ . FIG. 11B shows ^two types (B → C → D and B → A ⁾ of elements π _BD ⁽²⁾ of B rows and D columns in the matrix of Π ^{(2) obtained} by squaring the matrix of Π ^(1). → (D) indicates the number of directed paths, and FIG. 11 (C) is an element π _BD ⁽³⁾ of B rows and D columns in the matrix of Π ⁽³⁾ obtained by cubeing the matrix of Π ^(1). The number of directed paths on the street (B → C → A → D) is shown.

  Returning to FIG. 6, in step 64, in order to raise the matrix up to the set maximum path length LMAX, it is determined that path length <maximum path length. When the path length is less than the maximum path length, the matrix power is further increased so as to increase the path length by 1 in step 65, and an evaluation path number list that is an element of the matrix in the obtained evaluation path length is added. Run again. If route length ≧ maximum route length, the process proceeds to step 66, and the evaluation route number list corresponding to the evaluation route length is moved to the next process. Here, the maximum route length is an integer of 1 to the number of participants. For example, by setting the maximum route length to be considered to LMAX, the directional route that follows more participants than LMAX is considered to be unlikely to be collusion, so it is removed from consideration of the collusion degree described later, Will have no effect. Thereby, the calculation time of collusion degree calculation can be shortened.

  As a result of the above, a directed graph and accompanying matrix are created at the directed edge between the evaluator and the evaluator nodes, and the directed path length is determined from the matrix element of the directed path length up to the maximum path length corresponding to the power of the matrix. You can get a number.

  The above is the description of the evaluation route search unit 12 including the evaluation route length acquisition unit and the evaluation route number acquisition unit 14 of FIG. Next, the reputation information calculation means 13 will be described. The reputation information calculation means 13 calculates the collusion between each participant according to the directed path length and the number of directional paths based on the evaluation path number list acquired by the evaluation path search means 12, and determines the collusion degree. Accordingly, the evaluation value is reduced and the reputation value is calculated. The reputation information calculation means 13 includes a collusion degree calculation means 15 for obtaining a collusion degree from the evaluation path number list, an evaluation value correction means 16 for correcting an evaluation value from the evaluation information input means 10 based on the collusion degree, and a corrected evaluation. It comprises reputation value calculation means 17 for collecting values and calculating a reputation value.

  Here, the reputation information calculation means 14 is demonstrated based on the flowchart shown in FIG. It is desirable that the reputation information calculation by the reputation information calculation means 14 is basically performed for all evaluation values based on the evaluation route search every time the evaluation information is input by the evaluation information input means 10. Note that this reputation information calculation may be performed by batch processing at regular intervals in order to reduce the amount of calculation.

The examples of the participants A to E described in FIG. 7 to FIG. 11 are expanded, and as a general expression, a matrix for the evaluation path length k (power k) (k = 1... LMAX) in the participant ij. When the evaluation path number list including the elements π _ij ^(k) is obtained, the collusion degree is as follows. That is, in calculating the degree of collusion between participants according to the directed path length and the number of directed paths in step 121 in FIG. 12, the degree of collusion between participant i and participant j is determined by the number of matrix elements π _It becomes large depending on _ij ^(k) , and it is considered that the possibility of collusion is low when the directed path length is long. Therefore, the collusion degree π _ij of the evaluator i and the evaluator j calculated by the collusion degree calculation means 15 is small when the directed path length is long (k is large), and the number of matrix elements is large (π _ij ^{( Considering} that ^k) is large), it can be expressed as, for example, the following equation.
π _ij = Σ 2 ^−k × π _ij ^(k) (k = 1... LMAX)

In step 122, the collusion degree obtained by the collusion degree calculation means 15 is reflected in the evaluation value input by the evaluation information input means 10. That is, the evaluation value correction means 16 considers the collusion degree calculated in step 121 and corrects the evaluation value so that the evaluation becomes lower as the collusion degree is higher. For example, when the evaluation value of each evaluator i and evaluator j recorded by the evaluation information recording means 19 is e _ij , it depends on the back link from the evaluator j to the evaluator i, that is, the collusion degree π _ji The collusion degree corrected evaluation value e ′ _ij corrected in this way is expressed by the following equation as an example.

Next, in step 123, a reputation value is calculated from the collusion degree corrected evaluation value _e'ij . That is, the reputation value calculation means 17 updates the reputation value data using the calculated collusion degree corrected evaluation value e ′ _ij . In this example, the sum of evaluation values e ′ _ij corrected for collusion from other evaluators is considered as a reputation value. That is, the reputation value of a certain participant x is obtained as the sum of evaluation values e ′ _ix corrected for the collusion degree from all other participants (m people) by the following equation.
Σe ' _ix (i = 1 ... m)

In step 124, the reputation value of the participant x stored in the reputation information recording means 20 is updated by using the calculated reputation value Σe ′ _ix as a new reputation value.
As described above, the reputation information calculation means 13 obtains the collusion degree, corrects the input evaluation value by this collusion degree, and calculates the reputation value by summing up the corrected evaluation values. The reputation information of the participant (evaluated person) recorded in the information recording means 20 is updated.

FIG. 13 is an operation flowchart of the reputation information output means 18. When a reputation information acquisition request input from the participant's user terminal via the Internet or the like is received (step 131), the reputation information acquisition request includes a user ID for which reputation information is to be acquired. Therefore, the stored data in the reputation information recording means 20 is searched using this user ID as a key (step 132), and the corresponding reputation value is obtained. The obtained reputation value is transmitted to the participant's user terminal using communication means such as the Internet (step 133). The reputation information in the reputation information recording means 20 obtained at this time is a reputation value corresponding to the user ID of the person to be evaluated as shown in FIG.

In addition, although the description so far has used the directed path | route about the directed graph, this directed path | route is the same as the evaluation path | route of an evaluator and a to-be-evaluated person. In addition, according to the present embodiment, the reputation information becomes a global value without depending on the cluster to which the individual belongs, as in the conventional system, so that it is not disadvantageous to the initial user, and the reputation score value is set. Accordingly, it is possible to calculate an easy-to-use reputation value for the purpose of discounting membership fees or distributing incentives.
In the above description, the description has been given based on the hardware shown in FIG. 1, but this embodiment can also be configured by a computer program, and thus the computer program can be recorded on a recording medium.

It is a block diagram which shows the reputation information calculation apparatus in embodiment of this invention. It is an operation | movement flowchart of the reputation information calculation apparatus with respect to a request. It is a relationship diagram which illustrates the evaluation information which an evaluation information recording means manages. It is a relationship diagram which illustrates the reputation information which a reputation information recording means manages. It is an operation | movement flowchart by an evaluation information input means. It is an operation | movement flowchart by an evaluation route search means. It is explanatory drawing of a directed graph and a matrix based on it. It is explanatory drawing of the original matrix based on a directed graph. It is explanatory drawing of the matrix which squared the matrix of FIG. It is explanatory drawing of the example of a partial calculation of a matrix element. It is explanatory drawing which shows the directed path length between BDs, and the number of directed paths. It is an operation | movement flowchart by a reputation information calculation means. It is an operation | movement flowchart by a reputation information output means. It is explanatory drawing of the commercial transaction on the internet.

Claims (8)

The community formed by the network, the evaluator who is the participant of the community to be evaluated and the evaluated person who is the participant of the community to be evaluated In the reputation information calculation device for calculating the reputation information of the evaluated person by totaling the evaluation information,
Evaluation information input means for recording evaluation information received from the evaluator's terminal in the evaluation information recording means;
Based on a directed graph that connects the presence of evaluation information for the evaluated person by the evaluator with a directed edge using the community participants as nodes, the evaluation information includes an evaluation path starting from the evaluated person and ending at the evaluator. Evaluation route searching means for searching by recording means;
Reputation information calculation means for calculating a collusion degree of evaluation information according to a search result of the evaluation route, and calculating reputation information of the evaluated person based on the collusion degree and the evaluation information,
Reputation information output means for outputting reputation information from the reputation information recording means in which the reputation information is recorded;
A reputation information calculation apparatus comprising:   In the reputation information calculation apparatus according to claim 1, the evaluation path search means configures a matrix in which the evaluator and the evaluated person are divided into rows and columns and the presence or absence of a directed side is associated with an element, A reputation information calculation device characterized by specifying an evaluation path by taking a power of a matrix. In the reputation information calculation apparatus according to claim 1, the evaluation route search means includes:
The evaluation path length acquisition means for acquiring the length of the evaluation path from the evaluated person to the evaluator by the connection of the directed edges constituting the evaluation path, and the reputation information calculation means includes the evaluation path length acquisition means. A reputation information calculation device characterized by obtaining a collusion degree corresponding to the length of the evaluation path obtained by the above.   The reputation information calculation apparatus according to claim 1 or 3, wherein the evaluation path search means includes evaluation path number acquisition means for acquiring the number of different evaluation paths from the same evaluated person to the same evaluator. The reputation information calculation unit obtains a collusion degree corresponding to the number of evaluation paths obtained by the evaluation path number acquisition unit.   5. The reputation information calculation device according to claim 1, wherein the reputation information calculation means obtains a collusion degree by accumulating the number of evaluation routes weighted according to the length of the evaluation route. Reputation information calculation device characterized by having means.   6. The reputation information calculation apparatus according to claim 1, wherein the reputation information calculation unit includes an evaluation value correction unit that corrects an evaluation value of an evaluated person based on a collusion degree, and the evaluation value correction unit. And a reputation value calculation means for accumulating the corrected evaluation value according to the number of evaluations received by the person to be evaluated.   A program that causes a computer to function as the reputation information calculation apparatus according to claim 1.   A computer-readable recording medium on which the program according to claim 7 is recorded.

Images