JP3315752B2 - Similarity table creation device in concept hierarchy - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for creating a similarity table in a concept hierarchy, which creates a similarity table representing the similarity between attribute values formed in a hierarchy.

[0002]

2. Description of the Related Art In information processing such as information retrieval, various types of data are expressed as pairs (cases) of attributes and attribute values, and it is often necessary to evaluate the similarity between attribute values. Such a similarity determination can be made based on a weight of an attribute or a similarity between attribute values, in which a similarity is previously set between attribute values. Here, the similarity is a measure for evaluating the similarity between attribute values. The similarity between attribute values can be determined using this similarity, for example,

## EQU00001 ## Similarity r = Σ (weight of attribute w.times.similarity d between attribute values).

A set of similarities between attribute values of a certain attribute is called a similarity table. This similarity table can be created by setting a similarity for each set of attribute values.

[0004]

When the relationship between the attribute values is parallel, setting or changing one similarity does not affect other similarities in principle. There is almost no need to consider the balance and consistency of the entire similarity table.

However, each attribute value may have a tree-like hierarchical relationship in which the superordinate concept and the subordinate concept depend on each other. The hierarchy constituted by each attribute value is called a concept hierarchy. In the concept hierarchy, all attribute values are linked by a tree structure starting from a single root. For this reason, when setting the similarity between attribute values in the concept hierarchy, it is necessary to consider the balance and consistency of the entire concept hierarchy. That is, in the concept hierarchy, if the similarity of the upper concept level of the hierarchical structure is changed, the similarity of the lower concept level dependent on the higher concept is naturally affected, and conversely, the similarity of the lower concept is changed. Then, the similarity between the higher-level concept and the lower-level concept that is dependent on the lower-level concept is also affected by the similarity.

For this reason, each time the user sets or changes the similarity, the user considers the spread to other similarities and adjusts the entire concept table in order to maintain the balance and consistency of the similarities in the entire hierarchical structure. That was a complicated procedure. Therefore, there has been a problem that not only the efficiency and accuracy of creating the concept table is reduced, but also it becomes difficult to reflect the user's intention in the concept table.

Here, when the similarity is set, AHP (analytic hierarchy pro- gram) which is one of the decision making techniques is used.
cess). (Reference: Saaty,
TL: The Analytic Hierarchy Process, 1980). This method is one of the representative methods of Operations Research, and its features are to classify the evaluation criteria into several stages, and for each of the evaluation criteria, focus on each of the evaluation targets by paired comparison. Is evaluated, the overall weight is calculated based on the point, and the order of importance of the evaluation target is determined.

However, in this method, when the pair comparison is not consistent, it is necessary to redo the pair comparison.
A heavy burden is imposed on the user. In addition, the similarity is set for each set of attribute values, and the similarity in the opposite direction to one attribute value is not necessarily the same as that in the other direction. There was a problem that sex was limited to one direction.

The present invention has been proposed to solve the above-mentioned problems of the prior art. The object of the present invention is to maintain the balance and consistency of similarity in the entire concept hierarchy while maintaining the attribute values. It is an object of the present invention to provide a similarity table creating apparatus for creating a similarity table between concepts according to a user's intention.

[0010]

In order to achieve the above object, a similarity table creating apparatus in a concept hierarchy according to the first aspect of the present invention includes a similarity table representing each similarity between respective attribute values hierarchically arranged. A similarity table storage unit for storing, a constraint input unit for inputting a constraint to be satisfied by each of the similarities, and a constraint storage unit for storing at least the constraint input from the constraint input unit. A similarity input unit that inputs a new similarity to be set in the similarity table; a similarity setting unit that sets the new similarity in the similarity table; and the similarity setting unit that sets the new similarity based on the constraint. Having similarity propagation means for propagating to each of the similarities stored in the similarity table, and matching adjusting means for adjusting each of the similarities in the similarity table based on the constraint; And it features.

According to a second aspect of the present invention, there is provided the apparatus for creating a similarity table in the concept hierarchy according to the first aspect, further comprising a propagation mode selecting means for selecting the mode of the propagation.

According to a third aspect of the present invention, in the concept similarity table creating apparatus of the first aspect, an initial value for setting a predetermined default value as an initial value of each similarity in the similarity table. It is characterized by comprising setting means.

[0013]

The present invention having the above configuration has the following operation. That is, in the invention of claim 1, the similarity table storage stores the similarity between the attribute values, and the constraint input from the constraint input unit is stored in the constraint storage. When the user inputs a new similarity to be set between desired attribute values from the similarity input means, the similarity setting means sets a new similarity in a similarity table, and the similarity propagation means The new similarity is propagated to the similarity stored in the similarity table based on the constraint. Depending on the specific mode of the propagation, a violation of the constraint (ie, infringement, in other words, imbalance or mismatch) may occur between the similarity within the propagation range and the similarity outside the propagation range. But in such a case,
The matching adjustment unit adjusts each similarity in the similarity table based on the constraint.

As described above, according to the first aspect of the present invention, the propagation of the new similarity is performed based on the constraint including the constraint input by the user, so that the user's intention is reflected in the similarity table. In addition, since the similarities are adjusted based on the constraints, the balance and consistency of the similarities in the entire concept hierarchy are maintained.

According to the second aspect of the present invention, since the user can select the propagation mode by the propagation mode selection means, it is possible to obtain a similarity table suitable for the purpose of use.

In the invention according to claim 3, the initial value setting means sets a predetermined default value as an initial value of each similarity in the similarity table. You are free from the complicated procedure of entering values.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an apparatus for creating a similarity table in a concept hierarchy (hereinafter, referred to as "the present apparatus") according to an embodiment of the present invention will be specifically described with reference to the drawings. The present device is realized on a computer, and each function of the present device is realized by operating the computer according to a predetermined procedure expressed in a program format. Therefore, the present device will be described below assuming a virtual circuit block having each function of the present device.

(1) Configuration of the Embodiment FIG. 1 [Schematic Configuration] First, FIG. 1 is a block diagram showing the configuration of the present apparatus. That is, the present device includes a keyboard 1 for input, a display device 2 for output, and a processing unit 3 for processing information. The processing unit 3 includes an I / O control circuit 4 and a concept hierarchy file 5 for storing a concept hierarchy hierarchically constituted by attribute values. The apparatus further includes a similarity storage file 6 for storing a similarity table representing the similarity between the attribute values, a similarity temporary file 7 for storing the similarity table being created,
A file management unit that transfers a similarity table between the similarity storage file and the similarity temporary file; The similarity storage file 6, the temporary similarity file 7, and the file management unit 8 constitute the similarity table storage unit.

Also, the present apparatus has a constraint input unit 9 for taking in from the keyboard 1 constraints that should be satisfied by the similarities.
And a constraint file 10 (corresponding to the constraint storage means) for storing at least the constraint input from the constraint input unit 9. In addition, keyboard 1,
The I / O control circuit 4 and the constraint input unit 9 constitute the constraint input means.

The apparatus further comprises a similarity input unit 11 for inputting a new similarity to be set between the attribute values from the keyboard 1 and a similarity setting unit for setting the new similarity in the similarity table. 12 and a similarity propagating unit 13 (corresponding to the similarity setting unit) that propagates the new similarity to the similarity stored in the similarity table based on the constraint (the similarity propagating unit). Equivalent). The keyboard 1, the I / O control circuit 4, and the similarity input unit 11 constitute the similarity input means. In addition, the present apparatus includes a matching adjustment unit 14 (corresponding to the matching adjustment unit) that adjusts each similarity of the similarity table based on the constraint, and notifies a similarity that cannot be adjusted. ing.

The apparatus further comprises an initial value setting unit 15 (corresponding to the initial value setting means) for setting an initial value of each similarity, and a propagation mode for taking the mode of propagation from the keyboard 1. An input unit 16 (corresponding to the mode selection unit), a display control unit 17 for displaying information such as the content of the similarity table on the display device 2, and an instruction for capturing the display-related instruction from the keyboard 1 And an input unit 18.

[Concept Hierarchy] The concept hierarchy in this embodiment is stored in the concept hierarchy file 5 in advance.
The content is a simple model as shown in FIG. In this specification, in each of the drawings showing the structure of the concept hierarchy, a straight line indicates a dependency relationship between attribute values of a superordinate concept and a subordinate concept, and a bidirectional arrow indicates a degree of similarity set between attribute values. . That is, in this conceptual hierarchy, the root (root)
t) is dependent on the attribute values “spring” and “summer”, of which “spring” is dependent on the attribute values “March”, “April” and “May”, and further the attribute value is The attribute values “early March” and “late March” are subordinate to “March”. Also,
The attribute value "Summer" that depends on the route includes the attribute value "June",
"July" and "August" are dependent, and among them, the attribute value "August" is dependent on the attribute values "Early August", "Mid August" and "Late August".

In this concept hierarchy, the attribute value “spring” is a superordinate concept (parent) when viewed from the attribute values “March”, “April” and “May”. 3
“Month”, “April” and “May” are lower concepts (children) when viewed from the attribute value “Spring”.

[Similarity table] The similarity set between these attribute values takes a value from 0 (completely dissimilar) to 1 (very similar or identical) according to the degree of similarity. Shall be rounded off to the fourth decimal place. A set of these similarities is a similarity table.

[Constraints] The constraints in this embodiment are stored in the constraint file 10, and the system constraints which this apparatus normally has and which are automatically applied to any similarity table, and the specific processing fields And a user-defined constraint on the similarity table.

The system constraint is derived from the nature of the concept hierarchy. Specifically, the similarity between one attribute value, which is a superordinate concept, and another attribute value is determined for each subordinate concept of the one attribute value. This is a constraint that the similarity between a certain attribute value and the other attribute value must be an average value (hereinafter, this constraint is referred to as an “average value rule”).

According to this average value rule, for example, FIG.
, The attribute values “March”, “April” and “5”
Month ”and the similarity between a certain attribute value x are a,
In the case of b and c, the similarity s between the attribute value “spring”, which is a superordinate concept of these attribute values “March”, “April” and “May”, and the x,

S = (a + b + c) / 3 must be satisfied.

Note that this average value rule is, in other words,
The similarity between an attribute value, which is a lower concept of one attribute value, and another attribute value is calculated by calculating the average value of these similarities as the similarity between the one attribute value, which is a higher concept, and the other attribute value. So that
It has to be set.

Incidentally, the system constraint may be a rule using a maximum value or a minimum value in the same manner as the above average value rule, depending on the nature of the processing target. That is, when the minimum value is used, if MAX () is a function that takes the maximum value of a plurality of argument values, the similarity s to x is:

S = MAX (a, b, c) must be satisfied. Further, when the minimum value is used, if MIN () is a function that takes the minimum value of a plurality of argument values, the similarity s to x is:

S = MIN (a, b, c) must be satisfied.

The average value rule is similarly applied to the case where the “other attribute value” is a subordinate concept of the “one attribute value” itself. For example, in FIG. 4, since the similarity between the attribute values “March” is 1, the attribute value “March”
And the similarity between the attribute value “April” and d, the similarity between the attribute value “March” and the attribute value “May” e, and the attribute value “April” and the attribute value “5”.
Assuming that the similarity of “month” is f, the attribute value “spring” and the attribute value “3”
Month s1

S1 = (q + d + e) / 3 = (1 + d + e) / 3.

The same applies to the case where the above-described maximum value and minimum value are adopted in the rule. When the constraint is determined based on the maximum value, the similarity s1 in FIG.

When s1 = MAX (q, d, e) = MAX (1, d, e) = 1 and the constraint is determined based on the minimum value, the similarity s1 in FIG.

S1 = MIN (q, d, e) = MIN (1, d, e) must be satisfied.

A user-defined constraint is a constraint specific to a domain such as a processing field or a processing target, and its contents are freely determined by a user. For example, a magnitude relationship of similarity between attribute values and an attribute are described. Constraints based on comparing similarities between values can be defined.

(2) Operation of the Embodiment: FIGS. 5 to 15 The creation of a similarity table in the present embodiment having the above configuration is performed as follows.

[Schematic Operation] FIG. 5 is a flowchart showing an example of a procedure for creating a similarity table by the present apparatus. In this procedure, first, the initial setting unit 15 performs initial setting (step 21). That is, in this initial setting,
The initial setting unit 15 creates an original form of the similarity table having a structure corresponding to the concept hierarchy in the concept hierarchy file 5 in the similarity temporary file 7 and assigns a predetermined default value to each similarity in the similarity table. Is set as an initial value. Note that the user can input a desired user-defined constraint to the apparatus via the constraint input unit 9 before creating the similarity table.

Next, when the user inputs a new similarity to be set between desired attribute values via the similarity input unit 11 (step 22), the similarity setting unit 12 calculates the new similarity. The similarity between the attribute values is set in the similarity table (step 23). When setting the new similarity, the matching adjustment unit 14 performs preliminary adjustment (step 2).
4). The preliminary adjustment is a process of adjusting the similarity of the surroundings so that the set new similarity and the surrounding similarity (another similarity serving as an application unit of the constraint) satisfy the constraint.
In this preliminary adjustment (step 24), when there is a similarity that cannot be completely adjusted and thus violates the constraint (step 25), the new similarity or the similarity around it has a new similarity. You will be asked again for input (Step 2
2).

If the infringement of the constraint is absent or resolved,
The user is inquired whether or not to change the propagation mode (step 26), and when the user inputs the change from the keyboard 1, the propagation mode input unit 16 receives an input (selection) of the propagation mode (step 27). Next, the similarity propagation unit 1
3 propagates the new similarity to each similarity in another similarity table based on the input propagation mode and constraint (step 28). Subsequently, the matching adjustment unit 14 performs the matching adjustment (Step 29). Here, the matching adjustment is a process of evaluating the consistency with the constraint for the entire concept hierarchy including the part that is not directly processed by the propagation, and adjusting each similarity to satisfy the constraint. In addition, the matching adjustment unit 14
If there is an infringement of the constraint caused by the inability to adjust completely (step 30), the user is notified of the infringement via the display control unit 17 (step 31), and the procedure returns to step 26. Note that the flowchart in FIG. 5 is an example, and when such infringement of the constraint is notified, the user may input the similarity again (step 26). The file management unit 8 copies and stores the similarity table in the similarity temporary file 7 to the similarity storage file 6 (step 32). Then, the user is inquired whether or not to end the creation (step 33). When the user inputs that the creation is to be ended, the procedure ends. The procedure from step 22) is repeated again.

[Initial Setting] In the initial setting (step 21), the initial value set for each similarity in the similarity table is a predetermined default value, and the default value of the similarity in this embodiment is the same. There is 1 between attribute values, between attribute values having a parent-child relationship, and between attribute values having a sibling relationship subordinate to the same attribute value, and between attribute values having no such relationship is 0.5.
For example, in the concept hierarchy of FIG.
Attribute values “March”, attribute values “April”, attribute value “5”
The similarity between “months” is 1 in each case.

As described above, the predetermined default value is set as the initial value of each similarity. As a result, the creation of the similarity table in the present embodiment is performed between each attribute value from the root to the lowest instance level of the concept hierarchy. Is started in a state in which a temporary similarity is set to. FIG. 6 is a diagram showing the contents of the entire similarity table in which a default value is set.

[Selection of Propagation Mode] The propagation mode in this embodiment is defined by three elements: the propagation direction, the propagation level, and the propagation principle. The user selects these elements in the propagation mode input in step 27. be able to. Here, the propagation direction includes downward (upper concept → lower concept), upward (lower concept → upper concept), and both directions.
Also, the propagation level is only one stage, up to the root level,
There is propagation to the instance level. Further, the propagation principles include ratio propagation, lateral level change, and "Undo" processing.

The “downward” in the propagation direction is defined as the similarity between the new similarity and the upper concept in the unit of applying the average value rule, and based on the new similarity, the similarity with the lower concept subordinate to the higher concept. The propagation direction that changes the degree. on the other hand,
“Upward” is a propagation direction in which the new similarity is the similarity with the lower concept in the application unit of the average value rule, and the similarity with the attribute value in the applicable unit is changed based on the new similarity. . Further, “both directions” refers to a case where both upward and downward propagation are performed.

In the propagation principle, “ratio propagation”
At the same ratio as the change ratio of the similarity changed by the new similarity, another similarity included in the constraint application unit is changed simultaneously with the new similarity. For example, according to the average value rule, in FIG. 3, the similarities included in the constraint application unit at the same time as the similarity s are the similarities a, b, and c.
In this case, the similarity is a = 0.40, b = 0.60, c = 0.80, s
When the similarity s is changed to 0.72 when the ratio is set to 0.60, and the ratio propagation is performed in the lower direction, the change in the similarity s (0.
The similarities a, b, c are also 1.2 according to the ratio 1.2 of 60 → 0.72).
The similarity is a = 0.48, b = 0.72, c = 0.9
6, s = 0.72.

The "horizontal level change" in the propagation principle means that, among the other similarities included in the application unit of the constraint at the same time as the new similarity, those that can be changed are restricted in relation to the superordinate concept. It is changed to satisfy. For example, according to the average value rule, in FIG. 3, similarities having a direct relationship with the similarity s in terms of constraints are similarities a, b, and c. And the similarity is a = 0.40, b = 0.60, c = 0.8
0, s = 0.60 and the similarity b, c is set to 0.72 when the change is suppressed as a user-defined constraint.
And the horizontal level is changed, the similarity becomes a = 0.7
6, b = 0.60, c = 0.80, s = 0.72. This principle is useful when the degree of similarity with the subordinate concept includes something that cannot be changed.

Here, a specific propagation mode based on a combination of the three elements of the propagation direction, the propagation level, and the propagation principle will be described. First, when a new similarity is set between two attribute values A and B in the conceptual hierarchy as shown in FIG. 7, the propagation direction of this similarity is downward (see FIGS. 8 and 9) from the input point. ) And above (FIGS. 10 and 1)
1) can be considered.

Of these, downward propagation can be considered in the following four ways.

(1-a) One-stage ratio propagation (1-b) Ratio propagation to instance level (1-c) One-stage horizontal level change (1-d) Horizontal level change to instance level Of these, up to instance level Propagation (1-b), (1-d)
In, when a default value (similarity other than the similarity set by the user as a new similarity; the same applies hereinafter) is found in the propagation path, the propagation is stopped by changing the default value. This is because the default value is a tentatively set similarity, and the influence of the default value does not need to be applied to an instance level lower than this.

In the horizontal level changes (1-c) and (1-d), the default values are also subject to adjustment such as equal distribution in order to satisfy the higher similarity. Here, Table 1 shows a propagation process that can be selected for the presence or absence of a default value in the downward propagation.

[0047]

[Table 1] In this classification, ○ indicates a selectable combination, ×
The marks represent combinations that cannot be selected.

The following four types of upward propagation are conceivable.

(2-a) Ratio propagation only in one stage (2-b) Ratio propagation up to root level (2-c) Horizontal level change only in one stage (2-d) Horizontal level change up to root level Among these, ratio propagation (2 -a) and (2-b) cannot be selected when the similarity with the lower concept includes a default value and the similarity with the higher concept is a user setting value. This is because the upper-level user setting value has no meaning under the influence of the lower-level default value.

In the horizontal level change (2-c), (2-d), if the similarity with the superordinate concept is the default value and the similarity with the subordinate concept is all user set values, Can't choose. This is because, if the user setting value that should be prioritized over the default value is changed in order to maintain the default value, the similarity table becomes contrary to the user's intention.
Here, Table 2 shows propagation processes that can be selected for the presence or absence of a default value in the downward propagation.

[0051]

[Table 2] In the present apparatus, it is also possible to select “Undo” processing in which constraints such as ratio propagation and lateral level change are not propagated. Here, “Undo” changes the similarity with the higher concept in order to satisfy the average value rule while giving priority to the latest similarity input by the user.

For example, in FIG. 3, the similarity is a = 0.4.
If 0, b = 0.60, c = 0.80, and s = 0.60, the similarity a is 0.48, the similarity b is 0.72, and the similarity c is 0.96.
When the “Undo” process is performed, these values are prioritized, and the similarity s is changed based on the average value rule. The similarities are a = 0.48, b = 0.72, c = 0.96, s = It becomes 0.72.

[Similarity Table Display Format] In this apparatus, the similarity, which is the content of the similarity table, is displayed on the display device 2 via the display control unit 17. The display of the similarity is usually performed in the order of breakdown (expansion) from the top concept side to the lower concept side. The user can input the similarity table via the instruction input unit 18 regardless of this order. Can be designated as a display target. This display is performed in a format that indicates the directionality of the similarity. That is, the similarity of one attribute value to one attribute value is not necessarily the same as the similarity of the one attribute value to the other attribute value. For this reason, in the present embodiment, the display of another attribute value (hereinafter, referred to as “case”) with respect to the one attribute value (hereinafter, referred to as “problem”) is provided on the left side of the table in a vertical manner. A case column is provided horizontally above the intersection column. Tables 3, 4, and 5 are display examples of initial values of the similarity table.

[0054]

[Table 3]

[0055]

[Table 4]

[0056]

[Table 5]

[Example of Similarity Setting and Propagation] Next, an example of similarity table creation by the present apparatus will be described. That is, when the user sets new similarities shown in Table 6 with respect to the initial values shown in FIG. 6, as a result of the preliminary adjustment, the similarities are changed as shown in Table 7.

[0058]

[Table 6]

[0059]

[Table 7] At this time, there is no infringement of restrictions.

Also, in this case, even in the field where the value is the same before and after the setting, the value type is changed from the default value to the user setting value, and the user setting value has priority over the default value. Therefore, the similarity whose value type has been changed in this way is subsequently handled according to the type. When the ratio propagation is performed from this state downward (upper concept → lower concept) to the instance, the similarity has the contents shown in Table 8.

[0061]

[Table 8] Subsequently, the matching adjustment of the entire concept hierarchy is performed, and the similarity has the contents shown in Table 9.

[0062]

[Table 9] At this time, the entire similarity table has the contents shown in FIG.

Further, when the user sets a new similarity shown in Table 10, as a result of the preliminary adjustment, the similarity is changed as shown in Table 11.

[0064]

[Table 10]

[0065]

[Table 11] At this time, there is no infringement of restrictions.

Then, in both directions (upper concept → lower concept, lower concept → upper concept), the last (root, instance)
If the horizontal level is changed until then, the similarity has the contents shown in Table 12.

[0067]

[Table 12] Subsequently, the matching adjustment of the entire hierarchical structure is performed. In the horizontal level change, since the value is set so as to satisfy the similarity with the higher concept, the matching is satisfied. At this time, the entire similarity table has the contents shown in FIG.

Further, when the user sets the new similarities shown in Table 13, as a result of the preliminary adjustment, the similarities are changed as shown in Table 14.

[0069]

[Table 13]

[0070]

[Table 14] At this time, there is no infringement of restrictions.

When the “Undo” process is performed, the similarity has the contents shown in Table 15.

[0072]

[Table 15] Subsequently, the matching adjustment is performed on the entire hierarchical structure, and the similarity is as shown in Tables 16 and 17.

[0073]

[Table 16]

[0074]

[Table 17] At this time, the entire similarity table has the contents shown in FIG.

Further, when the user sets the new similarities shown in Table 18, as a result of the preliminary adjustment, the similarities are changed as shown in Table 19.

[0076]

[Table 18]

[0077]

[Table 19] At this time, there is no infringement of restrictions.

When the ratio propagation is performed only one stage upward (lower concept → higher concept), the similarity is as shown in Table 20.

[0079]

[Table 20] Here, as a result of the consistency adjustment of the entire hierarchical structure,
As shown in (2), a similarity (shaded display portion) that violates the consistency is detected.

In this case, the inconsistency is presented on the display device 2 by the following message. <Inconsistency message> ・ Similarity between problem “August” and case “March” is already 0.458
Is set to

From the inconsistency message, since the similarity between the problem “August” and the case “March” is a user setting value, the user can determine the similarity between the problem “August” and the case “May”. Therefore, it can be understood that it is not allowed to be a target of the ratio propagation. In this case, various measures may be taken by the user, such as switching the propagation principle to changing the horizontal level. In this case, if the user resets the new similarity shown in Table 21, the similarity becomes The results are shown in Table 22. At this time, the entire similarity table has the contents shown in FIG.

[0082]

[Table 21]

[0083]

[Table 22] (3) Effects of Embodiment As described above, according to this embodiment, propagation of the new similarity is
Since the determination is performed based on the constraints including the constraints input by the user, the intention of the user is reflected in the similarity table. In addition, since the similarities are adjusted based on the constraints, the balance and consistency of the similarities in the entire concept hierarchy are maintained. Further, in this embodiment, the user operates the propagation mode input unit 1
Since the propagation mode such as the propagation direction, the propagation level, and the propagation principle can be selected via the interface 6, a similarity table suitable for the purpose of use can be obtained. As described above, according to the present embodiment, each time the user sets or changes the similarity, the user considers the spread to other similarities, or maintains the balance and consistency of the similarities in the entire hierarchical structure. This eliminates the complicated procedure of adjusting the entire concept table.

In particular, in this embodiment, since the propagation range changes depending on whether the similarity in the propagation path is the default value or the user setting value, the user setting value is treated with priority over the default value. The user's intention is more reliably reflected in the similarity table.

In the present embodiment, the initial setting unit 15
Since a predetermined default value is set as the initial value of each similarity, the user is free from a complicated procedure of inputting the initial value of each similarity. That is, when the creation of the similarity table is started from a state where no similarity is set, the creation of the similarity table does not end unless the user sets the similarity between all the attribute values. On the other hand, in the present embodiment, the creation of the similarity table starts from a state in which a default value is set for each similarity from the root to the instance. Therefore, if the user sets only the new similarity between the attribute values determined to be necessary, the other similarities are determined by propagation, so that the similarity table can be created quickly.

(4) Other Embodiments The present invention is not limited to the above-described embodiments.
The following other embodiments are also included. For example, the concept hierarchy and the similarity may be recorded in a single file. Further, in the above embodiment, the default value is set as the initial value of the similarity, but the similarity table created in the past and stored in the similarity storage file 6 is copied to the temporary similarity file 7. May be used as an initial value. In this way, the improvement / update of the similarity table can be repeated.

The elements constituting the propagation mode are not limited to the propagation direction, propagation level, and propagation principle, but can be freely determined. Further, the similarity table creating apparatus in the concept hierarchy of the above embodiment is realized on a computer, but all or a part of the functions may be realized on a dedicated electronic circuit.

[0088]

As described above, according to the present invention, a similarity table between attribute values is created according to a user's intention while maintaining the balance and consistency of similarity in the entire concept hierarchy. Since the apparatus for creating a similarity table in a hierarchy can be provided, the creation of a similarity table becomes efficient and accurate, and the burden on the user is greatly reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing the contents of a concept hierarchy in an embodiment of the present invention.

FIG. 3 is a diagram showing a part of a concept hierarchy in the embodiment of the present invention.

FIG. 4 is a diagram showing a part of a concept hierarchy in the embodiment of the present invention.

FIG. 5 is a flowchart illustrating a procedure for creating a similarity table according to the embodiment of the present invention.

FIG. 6 is a diagram showing contents after setting initial values of a similarity table in the embodiment of the present invention.

FIG. 7 is a diagram showing a model of a part of the concept hierarchy in the embodiment of the present invention.

8 is a diagram (lower) illustrating a propagation direction of a new similarity set in the concept hierarchy of FIG. 7;

9 is a diagram (lower) illustrating a propagation direction of a new similarity set in the concept hierarchy of FIG. 7;

FIG. 10 is a diagram (upper direction) illustrating a propagation direction of a new similarity set in the concept hierarchy of FIG. 7;

11 is a diagram (upper side) illustrating a propagation direction of a new similarity set in the concept hierarchy of FIG. 7;

FIG. 12 is a diagram showing contents of a similarity table according to the embodiment of the present invention.

FIG. 13 is a view showing contents of a similarity table in the embodiment of the present invention.

FIG. 14 is a view showing contents of a similarity table in the embodiment of the present invention.

FIG. 15 is a diagram showing contents of a similarity table in the embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Keyboard 2 ... Display apparatus 3 ... Processing part 4 ... I / O control circuit 5 ... Concept hierarchy file 6 ... Similarity storage file 7 ... Similarity temporary file 8 ... File management part 9 ... Constraint input part 10 ... Constraint file 11 ... Similarity input unit 12 ... Similarity setting unit 13 ... Similarity propagation unit 14 ... Match adjustment unit 15 ... Initial setting unit 16 ... Propagation mode input unit 17 ... Display control unit 18 ... Instruction input unit

────────────────────────────────────────────────── ─── Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 17/30 G06F 9/44 JICST file (JOIS)

Claims (3)

(57) [Claims] 1. A similarity table storing means for storing a similarity table representing each similarity between attribute values formed in a hierarchy, and a constraint for inputting a constraint to be satisfied by each similarity. Constraint input means, constraint storage means for storing at least the constraint input from the constraint input means, similarity input means for inputting a new similarity to be set in the similarity table, and the new similarity A similarity setting unit configured to set the similarity table in the similarity table; a similarity propagation unit configured to propagate the new similarity to each of the similarities stored in the similarity table based on the constraint; A matching adjustment unit for adjusting each of the similarities in the similarity table by adjusting the similarity table. 2. The apparatus according to claim 1, further comprising a propagation mode selection unit for selecting the mode of propagation. 3. A similarity table creation in a concept hierarchy according to claim 1, further comprising an initial value setting means for setting a predetermined default value as an initial value of each similarity in the similarity table. apparatus.