JP2008209977A - Importance degree calculation device, importance degree calculation method, importance degree calculation program loaded with same method, and recording medium storing same program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To evaluate a target at an optimum state on the basis of a set of comparison evaluation elements. <P>SOLUTION: An input text including a comparison evaluation element composed of a character string indicting a target, a character string indicating a comparison target to be compared with the target, a character string indicating an attribute which becomes an evaluation item for evaluating the target and the comparing target, and a character string indicating the polarity of evaluation of the target is acquired and the comparison evaluation element is input (1). Then, two targets included in the comparison evaluation element are compared with each other, the number of times of supporting one target as compared with the other target is calculated and the calculated number of times is regarded as the number of supports (3). Then the targets are regarded as nodes, the relative merits of the two targets are regarded as the weight of an edge, and a transition probability matrix expressing a graph based on a specific model between the targets is generated on the basis of the number of supports (5). Thus, eigenvector is obtained which belongs to the maximum eigenvalue of the generated transition probability matrix, evaluation value is calculated on the basis of the eigenvector and the evaluation value is regarded as an importance degree. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a technique for calculating the importance of an object based on a comparative evaluation element included in a text.

  In general, the importance calculation technique for calculating the importance related to a matter has been mainly performed on a Web page having a hyperlink structure. In this technical field, a technique for creating a graph space in which a Web page is associated with a graph node and a hyperlink is associated with an edge of the graph, and the importance of each page is obtained based on the graph space (for example, Non-Patent Document 1, Non-Patent Document 2) is the mainstream.

In the decision-making field, there is a technique (for example, refer to Non-Patent Document 3) that obtains an evaluation value of a certain object in the whole based on superiority or inferiority relationship between two objects. Non-Patent Document 3 is a technique called an AHP (Analytic Hierarchy Process) method, which is a decision-making technique based on a paired comparison method. The AHP method evaluates superiority or inferiority between two subjects (a combination between all two possible subjects when applied to a plurality of subjects) in several stages, and calculates the importance of each subject based on the assessment. Is the method.
L. Page, S.M. Brin, R .; Motwani, T .; Wingrad, “The PageRank Citration Ranking: Bringing Order to the Web”, [online], [January 17, 2007 search], Internet <URL: http: // dbpubs. Stanford. edu: 8090 / pub / showDoc. Fulltext? lang = en & doc = 1999-66 & format = pdf & compression = & name = 1999-66. pdf>. J. et al. M.M. Kleenberg, “Authoritative sources in a hyperlinked environment”, “Journal of the ACM”, [online], ACM, [January 17, 2007 search], Internet <URL: http: // www. home / kleinber / auth.pdf>. Hitoshi Kigure, “Introduction to AHP”, [online], April 16, 2005, Hitoshi Kigure, [Search January 17, 2007], Internet <URL: http: // www. kogures. com / hitoshi / webtext / or-ahp / index. html>.

  Although it is known that there is a method as described above, there is no method for optimally evaluating an object from a set of comparative evaluation elements.

  PageRank (see Non-Patent Document 1) is an algorithm for determining the importance of a Web page. For example, a web page is regarded as a node, a hyperlink from web page A to web page B is regarded as a directed edge from node A to node B, and the web space is projected as a graph.

  In the model in this algorithm, all hyperlinks on the Web space are treated equally. That is, when i outlinks are output from a certain page (node), a 1 / i transition probability is assigned to all the outlinks. Therefore, the graph structure is not suitable for expressing superiority or inferiority between nodes having continuous values indicating which is better. Non-Patent Document 2 also uses a similar graph structure.

  The AHP method in Non-Patent Document 3 is a decision making method based on a paired comparison method. This is a technique for evaluating the superiority or inferiority between two objects (a combination between all two possible objects when applied to a plurality of objects) in several stages, and obtaining the importance of each object based on the evaluation. For example, in the AHP method, a matrix X as shown below is created.

The meaning of each numerical value is as follows.
X _ij = 5 → more of xj is better than xi X _ij = 3 → more of xj is slightly better than xi X _ij = 1 → xi and xj is X _ji for X _ij's the same extent is the X _ij = 1 / X _ji .
X _ji = 1/5 → xj is better than xi (xi is worse than xj)
X _ji = 1/3 → xj is slightly better than xi (xi is slightly worse than xj)
X _ji = 1 → xi and xj are about the same (xj and xi are about the same)
Then, the evaluation values of the respective objects x1, x2, and x3 are obtained by obtaining eigenvectors belonging to the maximum eigenvalue of this matrix.

  However, the AHP method cannot evaluate an object when superiority information of two objects is given as a set. This is because there are elements in the set of superiority and inferiority information that B is better than A and elements that A is better than B, and there are multiple interpretations of superiority or inferiority of the two objects It is to do.

In the AHP method, even if the superiority information between two objects is aggregated and uniquely derived from a set of superiority and inferiority information by using some method, the superiority relationship between the object i and the object j is finally expressed as X _ij And 1 / X _ij need to be converted. The relationship between X _ij and its reciprocal 1 / X _ij in the AHP method is derived to facilitate the calculation from the matrix, and the superiority information of the two objects derived from the data set in this form is directly obtained. Cannot be expressed.

  The present invention has been made based on the above-described problem, and an importance calculation device, an importance calculation method for optimally evaluating an object based on a set of comparative evaluation elements, an importance calculation program that implements the method, and The object is to provide a recording medium storing the program.

  In order to solve the above problem, the invention according to claim 1 is characterized in that a character string indicating a target, a character string indicating a comparison target to be compared with the target, and evaluation of the target and the comparison target are performed. Text acquisition means for acquiring an input text including a comparative evaluation element composed of a character string indicating an attribute as an evaluation item, a character string indicating an evaluation regarding the target, and a character string indicating the polarity of the evaluation regarding the target An importance calculation device for calculating the importance of the object from the set of comparison evaluation elements, wherein the comparison is performed by acquiring input text including the comparison evaluation element from a text acquisition unit and inputting the comparison evaluation element The number of supports is calculated by comparing the evaluation element input means and the two objects included in the comparative evaluation element, calculating the number of times one object is supported by the other object, and regarding the number of times as the number of support. Means and said A transition probability matrix that represents a graph based on a specific transition model related to the object is generated based on the number of support, considering the elephant as a node and the superiority or inferiority between the two objects as the weight of the edge between the objects. And a graph generation unit that calculates an eigenvector belonging to the maximum eigenvalue of the generated transition probability matrix, calculates an evaluation value based on the eigenvector, and considers the evaluation value as an importance level. It is characterized by that.

  The invention described in claim 2 is characterized in that in the invention described in claim 1, there is provided comprehensive evaluation value calculating means for calculating a comprehensive evaluation value of the object from the evaluation value obtained for each attribute.

  The invention according to claim 3 is the invention according to claim 1 or 2, wherein at least information including the object and an evaluation value calculated for the object, or at least an evaluation value and an evaluation item calculated for the object and the object And a target importance level output means for outputting information including an attribute that is regarded as “a”.

  The invention according to claim 4 is a character string indicating a target, a character string indicating a comparison target to be compared with the target, and a character string indicating an attribute as an evaluation item when evaluating the target and the comparison target And a text acquisition means for acquiring an input text including a comparative evaluation element composed of a character string indicating evaluation regarding the target and a character string indicating the polarity of evaluation regarding the target, and a set of the comparative evaluation elements The importance calculation method used in the apparatus for calculating the importance of the target from the above, wherein the input element including the comparative evaluation element is acquired from the text acquisition means, and the comparative evaluation element input step of inputting the comparative evaluation element And comparing the two objects included in the comparative evaluation element, calculating the number of times that one object is supported over the other object, and considering the number of times as the number of support, versus A transition probability matrix representing a graph based on a specific transition model related to the object is generated based on the support count. A graph generation step, an eigenvector belonging to the maximum eigenvalue of the generated transition probability matrix, an evaluation value is calculated based on the eigenvector, and an evaluation value calculation step that regards the evaluation value as importance It is characterized by.

  The invention according to claim 5 is characterized in that, in the invention according to claim 4, there is provided a comprehensive evaluation value calculating step for calculating a comprehensive evaluation value of a target from an evaluation value obtained for each attribute.

  The invention according to claim 6 is the invention according to claim 4 or 5, wherein at least information including the object and an evaluation value calculated for the object, or at least an evaluation value and an evaluation item calculated for the object and the object And a target importance level output step for outputting information including the attribute that is regarded as.

  The invention according to claim 7 is an importance calculation program, wherein the importance calculation method according to any one of claims 4 to 6 is described as a computer program executable by a computer.

  The invention according to claim 8 is a recording medium, wherein the importance calculation method according to any one of claims 4 to 6 is described as a computer program executable by a computer, and the computer program is recorded. Features.

  According to the first and fourth aspects of the present invention, a transition probability matrix related to objects can be calculated based on the set of inputted comparative evaluation elements, and further, an evaluation value can be acquired based on the transition probability matrix.

  In the inventions according to the second and fifth aspects, the comprehensive evaluation value of the object can be acquired from the evaluation value obtained for each attribute.

  According to the third and sixth aspects of the invention, the target importance can be output.

  The invention according to claim 7 can describe the importance calculation method according to any one of claims 4 to 6 as a computer program.

  According to the eighth aspect of the present invention, a computer program that implements the importance calculation method according to any of the fourth to sixth aspects can be recorded on a recording medium.

  As described above, according to the first and fourth aspects of the present invention, the evaluation value is regarded as the importance level, and the target can be optimally and accurately evaluated based on the importance level.

  According to the second and fifth aspects of the present invention, the object can be comprehensively evaluated with high accuracy based on the comprehensive evaluation value.

  According to the fourth and sixth aspects of the invention, it is possible to perform an operation (for example, display) on the output target importance.

  According to the invention of claim 7, it is possible to provide a computer program in which the importance calculation method is implemented.

  According to invention of Claim 8, the recording medium which recorded the computer program which mounted the importance calculation method can be provided.

  These can contribute to the data analysis technology field.

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

[First Embodiment]
The importance calculation apparatus according to the first embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing a device configuration of the importance calculation device according to the first embodiment. The importance calculation device according to the first embodiment is configured by those indicated by reference numerals 1 to 8. 1 includes a general data recording device (for example, a memory or a hard disk device). For example, the calculation result or the like may be recorded in a specific area of a memory provided in advance in the importance calculation device.

  What is indicated by reference numeral 1 in FIG. 1 is a comparative evaluation element input means. This comparison evaluation element input means 1 is a character string indicating the target, the comparison target to be compared with the target, the attribute as the evaluation item when evaluating the target and the comparison target, the evaluation with respect to the target, and the polarity of the evaluation with respect to the target Means for inputting a comparative evaluation element comprising, for example, a keyboard, an OCR (Optical Character Reader), a pen input, a voice recognition device, a means for reading a text file placed on the network, etc. Anything that can input text is acceptable. That is, the comparative evaluation element input unit 1 may acquire a text to be processed from a text acquisition unit provided in advance. For example, text may be acquired from a storage device provided in advance, and the text may be regarded as input text.

  What is indicated by reference numeral 2 is an input data recording means, in which the comparative evaluation element input by the comparative evaluation element input means 1 is recorded. The input data recording means 2 records data having a data structure such that each element constituting the comparative evaluation element can be discriminated like a database. An example of data stored in the data recording apparatus will be described with reference to FIG. One data of the comparative evaluation element data T1 stored in the data recording device is composed of a target CL11 which is a data element, a comparison target CL12, an attribute CL13 indicating an attribute to be compared, an evaluation CL14, and a polarity CL15. For example, the target CL11 has a value of “camera A”. The polarity is determined based on “evaluation” and has a value corresponding to either positive or negative. In FIG. 2, “P” indicates affirmative and “N” indicates negative.

What is indicated by reference numeral 3 is a support number calculation means. Based on the data recorded in the input data recording means 2, the respective support numbers between the target i and the target j are calculated. Note that the support number S _ij of the target i between the target i and the target j is the number of persons who describe the support condition that the target i is better than the target j. These support numbers are calculated for each attribute CL13 or without considering the attribute CL13. In addition, the support number can be calculated for the data recorded in the input data recording means 2 by using a technique such as an association rule.

  What is indicated by reference numeral 4 is a support number recording means. The support number recording means 4 is a means for recording the support number (data) calculated by the support number calculation means 3. An example of the data structure of the data to be recorded will be described with reference to FIGS. FIG. 3 shows an example in which the support number is calculated without considering the attribute, and FIG. 4 shows an example in which the support number is calculated in consideration of the attribute.

  One data in the support number data T2 in FIG. 3 includes a support condition CL21 and a support number CL22 which is a value corresponding to the support condition.

  One data in the support number data T3 in FIG. 4 includes a support condition CL31, an attribute CL32, and a support number CL33.

  What is indicated by reference numeral 5 is a graph generating means. The graph generation means 5 models the behavior of “a customer who moves (transitions) the sales floor in search of a better product (target)”, and generates a graph based on the model. That is, the graph is generated by regarding the object as a node and the superiority or inferiority between the two objects as the weight of the edge between the objects. Note that generating a graph is equivalent to generating a transition probability matrix.

  Hereinafter, an example of an action model of “a customer who moves in the sales floor in search of a better product” will be described.

The customer starts to see one product from the product group, and then sees other products one after another. How the customer will act next depends largely on the superiority or inferiority relationship between the product currently being viewed and other products. Next, the customer selects roughly from the following two options.
(Option 1) Stay in front of the product A that is currently being viewed and continue to view the product.
(Option 2) Move to the front of another product B.

  If there are relatively few products that are superior to the product A that is currently being viewed, the customer stays at that location and continues to view the product that is currently being viewed. Conversely, when there are many other excellent products, it is easy to move in front of other products.

  Next, description will be made on the behavior of the customer to select which product when the option 2 “moves before another product B”. The customer is more likely to select a product that is relatively supported than the product A currently being viewed as a product to be viewed next. That is, a person who states that “product A <product B (product B is supported more than product A)” is “product A> product B (product A is supported more than product B)”. The customer B selects the product B that is relatively more than the person described as the product to be viewed next.

  The graph generation means 5 in this embodiment generates a graph according to the above model. One object (product) is assumed to correspond to one node. The transition probability matrix A of this graph will be described below. In the following description, the total number of nodes (products) V is N. Further, unless otherwise noted, the target (product) is called a node.

A _ij is a transition probability from the node Vi to the node Vj, and i and j take values from 1 to N. These values may be calculated for each attribute based on the support count calculated in consideration of the attribute, or may be calculated based on the support count calculated without considering the attribute.

  The calculation formula of the transition probability matrix A is shown below.

  In addition, the meaning of the term used by the calculation formula of said transition probability matrix A is as follows.

  The number of supports is the number of elements stated that A is better than B between A and B.

The number of battles, for example, the number of battles of the target A is the total number of targets compared with the target A. Note that g _ij indicates whether or not there is an individual battle, for example, 1 when the node Vi played against the node Vj (with a battle), and 0 when the node Vi did not play against the node Vj (no battle). , And.

The number of wins is that the target A wins the target B when the support number of A is greater than the support number of B between the target A and the target B. The number of wins of the subject A is the number of times that the subject A has won another product. Note that w _ij is an individual game result, for example, 1 when the node Vi wins the node Vj and 0 when the node Vi loses the node Vj.

  The support number S is calculated from a set of evaluation elements.

  A calculation method for calculating the above numerical values will be described below.

The support number S is obtained by extracting an association rule from the input data recording means 2 storing the comparative evaluation element data T1. The formula for calculating S is shown below.
S _ij : Number of times the node Vi is supported by the node Vj S _ji : Number of times the node Vj is supported by the node Vi S _ij = Support (target = i, comparison target = j−> polarity = P) + Support (target = J, comparison target = i-> polarity = N)... Equation 6.1
S _ji = Support (target = i, comparison target = j-> polarity = N) + Support (target = j, comparison target = i-> polarity = P) Equation 6.2
W (i) and G (i) can be obtained from the respective support numbers S. In this way, the transition probability matrix A is created, and finally each element of the matrix is divided by the sum ΣAi of the rows including the element.
A _ij = A _ij / ΣAi ............ Formula 7
The sum of transition probabilities from each node is normalized to 1 by the above formula. Thus, a transition probability matrix A is obtained. A _ij represents the transition probability from the node Vi to the node Vj.

  What is indicated by reference numeral 6 is a transition probability matrix recording means. The transition probability matrix recording unit 6 stores the transition probability matrix generated by the graph generation unit 5.

What is indicated by reference numeral 7 is evaluation value calculation means. Based on the graph (transition probability matrix) generated by the graph generation means 5, the evaluation value S (V) of each node is obtained. In the present embodiment, the evaluation value S (V) is regarded as importance. The evaluation value obtained here corresponds to obtaining the probability that the customer who was looking at the product group (node group) finally stands in front of a certain product (is in the node). The formula for obtaining the evaluation value is shown below.
S (Vj) = (1-d) * [Sigma] ( _Aij * S (Vi)) + d (8)
Thus, the evaluation value S (Vj) of the node Vj is calculated from the evaluation value of the node having an edge on Vj. d is the probability that the user will move from a product (node) currently being viewed to a completely unrelated product (node). A certain initial value is given to all nodes, and the calculation by this formula is repeated. Specifically, this is equivalent to examining the repeated multiplication (multiplication) of the transition matrix A and obtaining the probability of the destination state, and is equivalent to the problem of obtaining the eigenvector belonging to the maximum eigenvalue of the transition probability matrix A. It becomes.

  In general, by repeating (multiplying) a matrix, the given matrix can be settled into a certain set of values. The stable combination of numerical values is called an eigenvector, and the scalar quantity characteristic of the eigenvector is called an eigenvalue. Such a calculation method is called eigenvalue decomposition.

  A final evaluation value is obtained by normalizing the finally obtained eigenvector to a probability vector (a vector in which the sum of elements is equal to 1).

  When a transition probability matrix is created for each attribute, a target evaluation value for each attribute is calculated. When created without considering the attribute, the evaluation value for each target is calculated without considering the attribute.

  Then, the result obtained by the evaluation value calculation means 7 is output to the detection result buffer 8. That is, the combination of “target name, evaluation value” is performed in the detection result buffer 8 when it is performed without considering the attribute, and the combination of “target name, attribute, evaluation value” is performed in the case of performing the attribute. Is output.

  More specifically, this embodiment will be described below.

  For example, it is assumed that the following result is obtained in the support number calculation means 3 when the attribute is not considered.

S _AB = 1 S _BA = 3 S _AC = 4 S _CA = 2
That is, in the above result, the number of elements that the object A is better than the object B is 1 and vice versa, and the number of elements that the object A is better than the object C is 4 Yes and vice versa. The matrix of the graph generated using these results is as follows.

  Then, each element is divided by the sum of each row vector.

The above matrix is the transition probability matrix A. A graph having this transition probability matrix A will be described with reference to FIG. In FIG. 5, the transition probability tr1 of the node V _A itself remains 6/19, the transition probability tr2 of the transition from the node V _A to V _B is 9/19, the transition probability tr3 of the node V _B itself remains 4/5, the node The transition probability tr4 for transition from V _B to V _A is 1/5, the transition probability tr5 for transition from node V _A to V _C is 4/19, and the transition probability tr6 for transition from node V _C to V _A is 1.

  Next, the evaluation value of each node is obtained based on this matrix. First, consider the following matrix transposed.

For example, the first row of this transpose matrix represents the weight of the link to V _A.

  Next, evaluation values S (A), S (B), and S (C) are calculated. These values are obtained by recursively obtaining each evaluation value as shown in the following equation. That is, the number of updates t is increased, t is incremented to infinity (t → ∞), and the finally obtained value is the evaluation value of the node.

  The history of the calculation is shown below. This is calculated by setting the initial value of each node to 1 and the probability of random jump to 3/10.

  This series of flows is equivalent to obtaining the eigenvector belonging to the maximum eigenvalue of the transition probability matrix A. This value is normalized to a probability vector (a three-dimensional non-negative vector in which the sum of elements is equal to 1). This normalized probability vector becomes an evaluation value. And this evaluation value can be regarded as importance.

[Second Embodiment]
The importance calculation apparatus according to the second embodiment will be described with reference to FIG. FIG. 6 is a block diagram showing the configuration of the second embodiment.

  6, comparative evaluation element input means 1, input data recording means 2, support number calculation means 3, support number recording means 4, graph generation means 5, transition probability matrix recording means 6, evaluation value calculation means 7, detection result buffer Eight components of 8 are the same as those in the first embodiment, and are the same as those having the same reference numerals in FIG. However, in the first embodiment, the second embodiment is performed only when an evaluation value is calculated for each attribute.

  In the second embodiment, a comprehensive evaluation value calculation means 9 and a detection result buffer 10 are newly added. These means will be described below.

The comprehensive evaluation value calculation means 9 is a means for calculating the comprehensive evaluation using the result of calculating the evaluation value for each attribute. Here, the comparison target is O ₁ ,..., O _s ,..., O _n , the attributes are X ₁ ,..., X _t , ..., X _m , and the evaluation value of the attribute X _t of the target O _s is V _st . . Based on this, the comprehensive evaluation value V _s of the target O _s is obtained by the following calculation formula.
V _S = P ₁ V _s1 + P ₂ V _s2 +... + P _m V _sm ............ Formula 11
However, P ₁ ,..., P _m are weights set for the respective attributes.

The detection result buffer 10 is a buffer for storing the result obtained by the comprehensive evaluation value calculation means 9. The data format of the stored data T4 is “target, target name, evaluation value”. For example, detection result data (evaluation value data for each attribute) T4 (for example, target CL41 = “product A”, attribute CL42 = “price”, evaluation value CL43 = “0.3”, as shown in FIG. )) Is stored. In this example, the evaluation value V _{A of} the product _A , the evaluation value V _{B of} the product _B , and the evaluation value V _C of the product C are calculated as the following equations. However, the weight for attribute CL42 = “price” is 0.7, and the weight for attribute CL42 = “design” is 0.3.
V _A = 0.7 0.3 + 0.3 0.5 = 0.36 ............ Formula 12.1
V _B = 0.7 0.5 + 0.3 0.1 = 0.38 Equation 12.2
V _C = 0.7 0.2 + 0.3 0.4 = 0.26 ………… Formula 12.3
A result like the above equation is stored in the detection result buffer 10 in the data format of detection result data (total evaluation value data) T5 in FIG. For example, the target CL51 = "product A" and the comprehensive evaluation value CL52 = "0.36" are stored.

  Here, an example of a procedure for calculating the transition probability matrix in the first and second embodiments will be described below.

  The procedure for calculating the transition probability matrix will be described with reference to FIG.

First, a comparative evaluation element set CV = {C ₁ , C ₂ ,..., C _m } is input (S101).

Next, a target appearing in the inputted comparative evaluation element set and a set of comparison targets V = {V ₁ ,..., V _N } are acquired (S102).

  Next, temporary variables i and j are each initialized to 1 (S103).

  Next, it is checked whether or not the temporary variable i is less than or equal to the repeated variable N (i ≦ N) (S104). If the temporary variable i is less than or equal to the repetition variable N, the process proceeds to step S105. If the temporary variable i is larger than the repetition variable N (i> N), the process is terminated.

  In step S105, it is checked whether or not the temporary variable j is less than or equal to the repeated variable N (j ≦ N). If the temporary variable j is less than or equal to the repetition variable N (j ≦ N), the process proceeds to step S106. If the temporary variable j is larger than the repetition variable N (j> N), the process proceeds to step S112.

  In step S106, it is checked whether or not the temporary variable i is equal to the temporary variable j. If the temporary variable i is equal to the temporary variable j, the process proceeds to step S110. If the temporary variable i is not equal to the temporary variable j, the process proceeds to step S107.

In step S107, "support number S _ij of V _i with respect to V _j is 0, and the support number S _ji of V _j for V _i is 0," the condition that it is checked whether or not satisfied. If the condition is satisfied, the process proceeds to step S108. If the condition is not satisfied, the process proceeds to step S109.

In step S108, 0 is substituted for the transition probability A _ij from V _i to V _j, and the process proceeds to step S111.

In step S109, obtains the transition probability A _ij from V _i from the support number S _ji of V _j with respect to the support number S _ij and V _i of V _i with respect to V _j to V _j (e.g., wherein "A _ij = .alpha.D _ji / ( _Calculated using (S _ji + S _ij ) ”), the process proceeds to step S111.

In step S110, the self-transition probability is calculated from the number of battles G (i) and the number of wins W (i) of Vi (for example, calculated using the formula “A _ij = βW (i) / G (i)”), Proceed to step S111.

  In step S111, the value of temporary variable j is incremented by 1, and the process returns to step S105.

  In step S112, the value of temporary variable i is incremented by 1, and the process returns to step S104.

  Note that the importance level calculation apparatus is a combination of “target name, evaluation value” calculated in the first and second embodiments as described above, or information combined with “target name, attribute, evaluation value”. May be provided with target importance output means for outputting. For example, the target importance output means outputs information combining the “target name, evaluation value” or information combining “target name, attribute, evaluation value” to the display means including the display device. , May be displayed.

  Each means provided in the importance calculation device may be controlled by control means (for example, a CPU (Central Processing Unit)) provided in advance in the importance calculation device.

  Also, the present invention can be realized by configuring a part or all of the functions of each means in the importance calculation devices of the first and second embodiments as a computer program and executing the program using the computer. Needless to say, the method (procedure) relating to the importance calculation device in the first and second embodiments can be configured by a computer program, and the program can be executed by the computer. A computer-readable recording medium (storage medium) such as FD (Floppy (registered trademark) Disk), MO (Magneto-Optical disk), ROM (Read Only Memory), memory card, CD (Compact Disk) , DV It can be recorded on a D (Digital Versatile Disk), a removable disk, etc., and can be stored or distributed. It is also possible to provide the above program through a network such as the Internet or electronic mail.

  Furthermore, a computer program describing a method related to the importance calculation apparatus described above may be implemented so as to access a memory or an external storage unit storing input / output data required for the method.

  As described above, in the first and second embodiments, an object is determined based on a set of comparative evaluation elements (evaluation object, comparison object, attribute, evaluation, polarity) obtained by comparing a plurality of objects mentioned by people. Can be evaluated accurately.

  Although the embodiments of the present invention have been described above, the present invention is not limited to the described embodiments, and various modifications can be made within the scope described in each claim.

  For example, the behavior of “a person who moves (transitions) in search of a place (object) that is easier to spend” may be modeled.

The block diagram which shows the structure of the importance calculation apparatus in 1st Embodiment. The figure which shows an example of the data stored in an input data recording means. The figure which shows an example of the support number data stored in a support number recording means when an attribute is not considered. The figure which shows an example of the support number data stored in a support number recording means when an attribute is considered. The figure which shows an example of the graph which has the produced | generated transition probability matrix. The block diagram which shows the structure of the importance calculation apparatus in 2nd Embodiment. The figure which shows an example of the evaluation value for every attribute stored in the detection result buffer. The figure which shows an example of the comprehensive evaluation value stored in the detection result buffer. The flowchart which shows an example of the procedure which calculates a transition probability matrix.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Comparison evaluation element input means 2 ... Input data recording means 3 ... Support number calculation means 4 ... Support number recording means 5 ... Graph generation means 6 ... Transition probability matrix recording means 7 ... Evaluation value calculation means 8 ... Detection result buffer 9 ... Comprehensive evaluation value calculation means 10 ... detection result buffer CL11 ... target CL12 ... comparison target CL13 ... attribute indicating the attribute to be compared CL14 ... evaluation CL15 ... polarity CL21, CL31 ... support condition CL22 ... support number CL32 ... attribute CL33 ... support number CL41, CL51 ... Target CL42 ... Attribute CL43 ... Evaluation value CL52 ... Comprehensive evaluation value T1 ... Comparison evaluation element data T2, T3 ... Supported number data T4, T5 ... Detection result data V _A , V _B , V _C ... Node tr1-tr6 ... Transition probability

Claims (8)

A string indicating the target,
A string indicating a comparison object to be compared against the object;
A character string indicating an attribute as an evaluation item when evaluating the target and the comparison target;
A string indicating the evaluation of the object;
A string indicating the polarity of the evaluation for the object;
A text acquisition means for acquiring an input text including a comparative evaluation element composed of: an importance calculation device for calculating the importance of the object from a set of the comparative evaluation elements,
A comparative evaluation element input means for acquiring an input text including the comparative evaluation element from the text acquisition means and inputting the comparative evaluation element;
Comparing two objects included in the comparative evaluation element, calculating the number of times that one object is supported over the other object, and assuming the number of times as a support number;
A transition probability matrix that represents a graph based on a specific transition model related to the target is regarded as a node, and the superiority or inferiority between the two targets is the weight of the edge between the targets, and based on the support count A graph generating means for generating;
An eigenvector belonging to the maximum eigenvalue of the generated transition probability matrix, an evaluation value is calculated based on the eigenvector, and an evaluation value calculating means that considers the evaluation value as importance;
An importance calculation device comprising: In the importance calculation apparatus according to claim 1,
A comprehensive evaluation value calculation means for calculating a comprehensive evaluation value of a target from evaluation values obtained for each attribute;
An importance calculation device comprising: In the importance calculation apparatus according to claim 1 or 2,
Information including at least a subject and an evaluation value calculated for the subject, or
An object importance output means for outputting information including at least an object, an evaluation value calculated for the object, and an attribute regarded as an evaluation item;
An importance calculation device comprising: A string indicating the target,
A string indicating a comparison object to be compared against the object;
A character string indicating an attribute as an evaluation item when evaluating the target and the comparison target;
A string indicating the evaluation of the object;
A string indicating the polarity of the evaluation for the object;
An importance calculation method for use in an apparatus for calculating the importance of the target from a set of comparison evaluation elements, comprising text acquisition means for acquiring an input text including a comparative evaluation element composed of:
A comparative evaluation element input step of acquiring an input text including the comparative evaluation element from the text acquisition means, and inputting the comparative evaluation element;
Comparing the two objects included in the comparative evaluation element, calculating the number of times that one object is supported over the other object, and considering the number of times as the number of support;
A transition probability matrix that represents a graph based on a specific transition model related to the target is regarded as a node, and the superiority or inferiority between the two targets is the weight of the edge between the targets, and based on the support count A graph generation step to generate;
An eigenvector belonging to the maximum eigenvalue of the generated transition probability matrix, calculating an evaluation value based on the eigenvector, and an evaluation value calculating step that considers the evaluation value as importance;
The importance calculation method characterized by having. In the importance calculation method according to claim 4,
A comprehensive evaluation value calculating step for calculating a comprehensive evaluation value of the target from evaluation values obtained for each attribute;
The importance calculation method characterized by having. In the importance calculation method according to claim 4 or 5,
Information including at least a subject and an evaluation value calculated for the subject, or
An object importance output step for outputting information including at least an object, an evaluation value calculated for the object, and an attribute regarded as an evaluation item;
The importance calculation method characterized by having.   7. An importance calculation program characterized in that the importance calculation method according to claim 4 is described as a computer program executable by a computer.   7. A recording medium, wherein the importance calculation method according to claim 4 is described as a computer program executable by a computer, and the computer program is recorded.

Images