JPH08249338A - Data base concept schemer integration support device - Google Patents

Abstract

PURPOSE: To provide a device which supports schemer integration that enables an integrated schemer designer to easily find a corresponding schemer element between object schemers and is troubled by neither classifications of a compli cated different variety nor respective coping methods. CONSTITUTION: Schemer information to be integrated is inputted and the attribute name of the inputted schemer is standardized according to a data item naming rule. It is converted into a concept graph and a term dictionary 107 which holds classification information on terms is used to find the similarity between both schemer elements of the schemer to be integrated. A user interface part 105 shows the calculated similarity between the schemer elements to the operator, and on the basis of the relation between the determined schemer elements, a schemer integration part 104 merges two concept graphs.

Description

Detailed Description of the Invention

[0001]

The present invention relates to an organization such as a company,
The present invention relates to an apparatus that supports the work of creating an integrated schema across a plurality of databases, which is required when using data accumulated in an existing database in a crosswise manner.

[0002]

2. Description of the Related Art When integrating conceptual schemas of databases, the schemas to be integrated are represented by the same data model (this is called a common data model), and the integration work is performed on the data model.

Conventionally, the entity-relationship model generally used to represent the conceptual schema has been used as it is as a common data model. In the case of schema integration using the entity-relationship model, the same concept may be represented by different schema expressions, so the procedure of finding the corresponding concept between the integration target schemas may be complicated, and schema integration support The device has not been developed.
In the past, research on how to classify the heterogeneity (the same concept is expressed in different forms) between the schemas to be integrated and how to deal with each case has been done.

[0004]

In the conventional schema integration by the entity-relationship model, the same concept may be represented by different schema expressions, such as the entity type on the one hand and the attribute on the other hand. is there.

FIG. 2 shows this, and shows an example of the schema of the entity relation model to be input. Reference numerals 201 and 202 in FIG. 2 denote schemas (represented by entity-relationship models) to be integrated. To represent the building that is the end of the communication line, in the schema 201 of FIG. 2 the attributes of the actual type “line” (“starting building”, “end building”) are used, and in the schema 202 of FIG. It is expressed as another entity type.

[0006] In the conventional technology, only a method for coping with the case where such a situation exists between schemas is prepared.
"Starting building" and "end building" as attributes on the schema 201 side, and "building" as the entity type on the schema 202 side
There was no specific way to find out from the enormous amount of schema information that is a corresponding schema element.

In the schema integration work, finding the corresponding schema element among the target schemas is performed by a comprehensive judgment of the designer of the integrated schema. As the scale of the schema became larger, the effort for this became enormous.

An object of the present invention is to solve such a conventional problem so that an integrated schema designer can easily find a corresponding schema element between target schemas, and
It is to provide a device that supports schema integration without being bothered by complicated classification of heterogeneity and countermeasures for each.

[0009]

The above objects of the present invention are achieved by an apparatus as described below. That is, by converting the schema to be integrated into a concept graph as a common data model and calculating the similarity between concepts in the concept graph, candidates for corresponding schema elements are narrowed down,
It is a device that increases the efficiency of work by presenting it to the integrated schema designer.

FIG. 1 is a diagram showing a configuration example of an apparatus for realizing the present invention. This device is, for example, as shown in FIG.
A schema information importing unit 100 for importing schema information to be integrated, a data standardizing unit 101 for standardizing attribute names of the imported schema according to a data item naming rule, and a schema to concept graph. The model conversion unit 102, the similarity calculation unit 103 that obtains the similarity between the schema elements of both the integration target schemas, and the calculated similarity of the schema elements are presented to the operator, and the operator shows the relationship between the schema elements. A user interface unit 105 for judging and confirming, and a schema integrating unit 104 for merging two concept graphs are provided. The basic word table 106 is used for standardization,
The term dictionary 107 holds classification information of terms used in the names of schema elements, and is used when calculating the degree of similarity.

[0011]

The problem to be solved by the present invention is solved as follows. FIG. 3 shows an outline of a procedure for performing schema integration using the schema integration support device of the present invention.

The data standardization unit 101 uses the schema information acquired by the schema information acquisition unit 100 as follows.
Whether or not the attribute name is standardized is determined, and if standardized, the name is standardized. If not standardized, the standardized name is standardized.

FIG. 4 is a data standardization of the attribute name of the schema information represented in the entity relation diagram of FIG. For example, in the schema 201 of FIG. 2, “starting point building” is changed to “starting point_building_name”.
Next, this entity relationship diagram is converted into a conceptual graph. The rules for conversion are according to FIGS. 5 and 6.

A schema 701 of FIG. 7 shows a result of converting the entity relation diagram of the schema 201 of FIG. 2 into a concept graph. Similarly, the schema 702 of FIG. 7 shows the result of data standardization of the entity relation diagram of the schema 202 of FIG. 2 and conversion into a concept graph.

Next, the similarity calculation unit 103 calculates the similarity between the schema elements. First, the name similarity calculation 322 shown in FIG. 3 calculates the similarity between “concepts” according to the name similarity. At this time, the term dictionary 107 is used to calculate the degree of similarity.

Since the names are not the same in terms of similarity only, the similarity is calculated by the peripheral similarity calculation 323 shown in FIG. 3 with the peripheral information of each "concept" taken into consideration. When the calculation of the peripheral similarity calculation 323 is performed for the schemas represented by the two conceptual graphs of the schema 701 of FIG. 7 and the schema 702 of FIG. 7, the schemas are converted in this way and the schemas are compared. So, in the conventional method, the calculation had to be performed in consideration of the difference such as "starting building", "end building" and "building".
In this method, the similarity between schema elements can be calculated by a simple brute force calculation.

Utilizing the similarity between the schema elements calculated above, a set of schema elements having a high similarity is sequentially presented to the schema integrator through the user interface unit 105. As a result, the schema integrator can easily obtain candidates for similar schema elements.

Based on the result of the degree of similarity, if the schema needs to be modified, it is modified (referred to as schema adjustment. Process 330 in FIG. 3), and the schema integration unit 104 is used to merge the concept graphs. As post-processing 350, the model conversion unit 102 is used to convert the merged concept graph into a model such as an entity-relationship model, if necessary.

The problems of the present invention are solved by the above.

[0020]

FIG. 3 shows a flow of schema integration work using the apparatus of the present invention. In FIG. 3, reference numeral 300 represents preprocessing for data standardization, 310 represents model conversion processing, and 320 represents schema comparison processing. Schema comparison process 320
Includes a name similarity calculation 322, a marginal similarity calculation 323, and a process 321 for finding a comparison target schema element including a process 324 for performing a total similarity calculation from them and a matching determination process 325. Reference numeral 330 represents schema adjustment processing, 340 represents schema merge processing, and 350 represents post-processing such as model conversion.

The data standardization unit 101 standardizes data. The data standardization unit 101 is described in JP-A-4-21518.
Using the "basic word table" (basic word table 106) as shown in "Data name assignment registration device" of Japanese Patent Publication No. 2), the basic words included in the data item name are matched, the modifiers included, Discover the main word and the demarcation word and make them the standard form of modifier + main word + demarcation word. If the classifier is missing, the classifier is estimated and added from the data type and value example of the data item. If both the main word and the segment word are missing, the entire word is regarded as the main word, and the segment word is added in the same manner as when the segment word is missing.

The model conversion unit 102 converts an entity-relationship model (ER model) whose name is standardized into a concept graph using the rules shown in FIGS. 5 and 6
In E, E is an entity type, R is a relation, M is a modifier, P is a main word, and C is a classifier. The converted result is shown in FIG.

Next, the similarity is calculated. According to FIG. 6, there are three patterns of concept names, "E", "P", and "P_C". Therefore, when obtaining the name similarity of the concept name, there are the following three ways.

(1) When both concept names are not decomposed (that is, "E" vs. "E", "P" vs. "P", "E")
Pair "P"). (2) When both concept names are decomposed (that is,
"P_C" vs. "P_C").

(3) When one of the concept names is not decomposed and the other is decomposed (that is, "P" vs. "P_"
C "," E "vs." P_C "). Here, in the case of (3), there is always a concept of "P" always in pair with "P_C" according to Fig. 6 (Fig. 6 (b)).
, It seems that P does not exist, but P =
Since it is E, there exists P). Therefore, "P" vs. "P_
"C", "E" vs. "P_C" need not be considered similar. Therefore, the similarity is 0 for this combination.
And The calculation method of the name similarity will be described below for the cases (1) and (2).

(1) When the concept name is not decomposed Whether the words are similar or not is disclosed in Japanese Patent Laid-Open No. 21518/1992.
A score is given depending on whether or not the words correspond to the same standard word in the "standard word correspondence table" in the "data name assignment registration device" of the publication No. 2.

Specifically, the algorithm shown in FIG. 8 is used. Two concept names to be compared are input (S10), and standard names corresponding to them are searched (S11, S14). If there is a standard name corresponding to both two concept names and they are equal, the name similarity is set to 1 (maximum value) (S18). If they are not equal, they are completely different and are set to 0 (minimum value) (S1
9). If only one standard name does not exist, or if neither standard name exists, the degree of similarity is obtained using a technique such as subsequence matching (S20).

An example of applying this algorithm is shown in FIG. The concept names “building” and “building” are both found as standard names in the standard word correspondence table, so the similarity is 1. On the other hand, "line" and "circuit"
Has a similarity of 0 because only one standard name is found and there is no substring matching.

(2) When concept names are decomposed Reference "Sekine, Kawashita, Machihara, Nakagawa: Data standardization method using terminology dictionary for systematic DB construction, IPSJ Transactions Vol. 34" The "similar data item classification function" in "No. 3 (1993)" is used to give a score depending on whether or not they are in the same classification.

Specifically, this is performed as shown in FIG. Two concept names to be compared are input (S30), whether the section words are the same (S31, S32), and whether the main words are the same (S
33, S34, S37, S38), and if both are equal, 1 is set as the similarity (S35). If only the classifiers are equal, the intermediate similarity N is calculated as the similarity.
2 is set (S36). On the other hand, when only the main words are equal, a higher degree of similarity N1 is set as compared with the case where only the segment words are equal (S39). If both are not equal,
The lowest similarity N3 is set. The degree of similarity between words that are not completely equal is calculated using a technique such as subsequence matching.

An example of applying this algorithm is shown in FIG. The concept names “line_code” and “line_number” do not have the same divisional word but the same main word, so the similarity N1 is set. On the other hand, for example, in the concept names “line_code” and “circuit_code”, the classification of the main words is not the same and the segment words are the same, so an intermediate similarity N2 smaller than N1 is set.

Next, using the name similarity calculated in this way, peripheral similarity is calculated. For concepts A and B, sets of concepts adjacent to each other are defined as SA and SB. The peripheral similarity is calculated by the following formula.

(Similarity around A and B) = sum (max
(Name similarity between elements in SA and SB) / min (number of SAs, number of SBs) where (number of SAs) ≤ (number of SBs), max is S
The maximum and sum in B represent the total sum for A (and vice versa).

Consider the cases of "line" (= A) and "circuit" (= B) in the example of FIG. It is assumed that these are not registered in the standard term dictionary 107. The two have low similarity in name.

These adjacent concepts are: SA = {line_code, line_speed, building} SB = {Circuit_code, circuit_speed, building} And min (number of SAs, number of SBs) = 3
Is.

When the name similarity between the concept of SB and the concept of SA is calculated, “circuit_code” is calculated as “line_code”.
“Code” is maximized (similarity is N2. N2 =
0.3), and for "Circuit_speed",
"Line_speed" is maximum (also 0.3),
For "Building", "Building" is the maximum (the value is 1).

Therefore, sum (max (name similarity between elements in SA and SB)) = 0.3 + 0.3 + 1.0 = 1.6. Therefore, (similarity around “line” and “circuit”) = 1.6 / 3≈0.53. A concept that is not considered to be similar only by a single name similarity is considered to be similar although its value is not high by considering peripheral similarities.

Finally, the weighted average of the similarity of the name and the similarity of the surroundings is set as the similarity. In the case of the above “line” and “circuit”, the name similarity is 0 and the peripheral similarity is 0.53. Therefore, if the name weight: peripheral weight is 1: 1, the final The similarity is (0 ² +0.53
²⁾ the ^1/2 / 2 ^1/2 ≒ 0.37.

Next, the user operates the user interface section 1
Use 05 to determine if the concepts match. A set of concepts whose similarity is higher than the threshold is presented to the user in descending order of similarity. Look at it and, if necessary, refer to its surroundings to establish agreement.

Further, the agreement of the relation between the concepts is also determined.
Then, the two concept graphs are merged. The merging algorithm is, for example, the document “Sowa, JF Conceptual Str.
uctures: Information Processing in Mind and Machin
e. Addison-Wesely. (1984) ”. The merged concept graph is converted into an entity-relationship model if necessary.

This conversion algorithm is shown in FIGS.
Shown in. org (C) represents the origin of the concept C. E is real
Body type, R is related, P is a main word, and C is a section word. under
The letters attached indicate the origin. For example, C_EIs
It is a concept that has a substantive type as its origin. Origin * is any origin
Is represented. C >> E means that the concept C is converted to the substantive type E.
It means that it was done. Basic idea of this conversion rule
Converts what was originally a substantive type to a substantive type,
The relationship between the concept converted to
The clerk is the rule to convert to the related type. [C₁]
-(R)-> [C ₂], Etc.
Follows the literature.

14 to 16 show examples in which the concept graph is converted into the entity relation model using this rule. Of FIG.
(a) and (b) show examples of two input schemas. The result of converting it into a concept graph is shown in Fig. 15 (a),
It is (b). The result of merging them is shown in (a) of FIG. The result of converting it into an entity-relationship model is shown in FIG.
(B) of.

[0043]

As described above, according to the database conceptual schema integration support device of the present invention, the efficiency of the work of integrating the conceptual schemas of the databases is improved.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of an apparatus that realizes the present invention.

FIG. 2 is a diagram showing an example of a schema of an entity relation model to be input.

FIG. 3 is a diagram showing an outline of processing of the present invention.

FIG. 4 is a diagram showing a result of standardizing data item names from the example of FIG.

FIG. 5 is a diagram showing a conversion rule from an entity relation model to a concept graph.

FIG. 6 is a diagram showing a conversion rule from an entity relation model to a concept graph.

FIG. 7 is a diagram showing a conceptual graph converted from the example of FIG.

FIG. 8 is a diagram showing a calculation process of a name similarity when the concept name is not decomposed.

FIG. 9 is a diagram showing an example of calculation of name similarity when concept names are not decomposed.

FIG. 10 is a diagram showing a name similarity calculation process when concept names are decomposed.

FIG. 11 is a diagram showing an example of calculation of a name similarity when concept names are decomposed.

FIG. 12 is a diagram showing an algorithm of model conversion.

FIG. 13 is a diagram showing an algorithm of model conversion.

FIG. 14 is a diagram showing an example of model conversion.

FIG. 15 is a diagram showing an example of model conversion.

FIG. 16 is a diagram showing an example of model conversion.

[Explanation of symbols]

 100 Schema Information Importing Part 101 Data Standardizing Part 102 Model Converting Part 103 Similarity Calculation Part between Schema Elements 104 Schema Integration Part 105 User Interface Part 106 Basic Word Table 107 Term Dictionary

Claims (1)

[Claims] 1. A schema information importing unit for importing schema information to be integrated in a database conceptual schema integration support device which supports inputting conceptual schema information of a plurality of databases and generating integrated schemas thereof. , Holds the data standardization part that standardizes the attribute names of the imported schema according to the specified data item naming rules, the model conversion part that converts the imported schema into a concept graph, and the classification information of the terms used in the names of schema elements The term dictionary, a similarity calculation unit that obtains the degree of similarity between schema elements of both integration target schemas by using the term dictionary, and the operator are presented with the calculated degree of similarity of the schema elements. A user interface part for judging and confirming relationships and merging of multiple concept graphs A database conceptual schema integration support apparatus comprising: a schema integration unit.