JPH06231188A - Device for sorting name of similar data - Google Patents

Abstract

PURPOSE:To sort the name of data from the view point of the similarity of data values. CONSTITUTION:Each entry of a dictionary 4 consists of a word number, the name of a word, segmented word flag, main word flag, and the word number (representitive word number) of the more generalized word of the word (representitive word). A dictionary check section 11 identifies the word which coincides with the name of the word in the dictionary at every word comprising the name of data inputted from an input device 2, and outputs the name of a word (the name of the representitive word) having the same value in the word number as the segmented word flag. main word flag, and the representitive word number of the word. An important word extraction section 12 extracts the segmented word and the main word from the name of data, extracting the representitive segmented word and the representitive main word. A sorting processing section 13 sorts the name of data by perform the sort processing with representitive segmented word as large sorting and with the representitive main word as small sorting and outputs the result from the output device 3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is designed to unify different data names given to data having the same intention in a database analysis and design work. The present invention relates to a similar data name classification device that identifies that data have the same intention.

[0002]

2. Description of the Related Art There are the following conventional techniques for classifying similar names. (Technology 1) A technique for decomposing a data name into terms and determining the similarity of data names from the similarity of each term has already been established (for example, Japanese Patent Laid-Open No. 3-41560). (Technology 2) A technology of sorting character strings to classify character strings having similar leading characters has already been put to practical use in commercial database management systems. In addition, the technique for decomposing data names into terms has already been reviewed, for example, by Shoichi Noguchi and Takenori Makino, "Natural Language Processing."
(Ohm, 1991), pages 27-34.

[0003]

An example of determining a data name from similarity and problems of the prior art will be described below. (Case 1) Similar data name classifications of “current customer number” and “new customer number” have similar values in that they are numbers related to customers, so it is necessary to determine similar data. (Case 2) Similar data name classification of “current customer number” and “current customer address” Since both have different values of number and address, it is necessary to determine that they are not similar. (Case 3) Similar data name classification of “current customer number” and “current order number” Since both are numbers for different entities, that is, customer and order, it is necessary to determine that they are not similar.

In the prior art, it can be determined that (case 1) is correct and similar, but (case 2) and (case 3) have the same data name even though they are not similar. Since many are included, there is a problem that they are determined to be similar.

An object of the present invention is to provide a similar data name classifying apparatus which can correctly determine the above (case 2) and (case 3) and can classify data names having similar data values. Especially.

[0006]

In order to achieve the above object, a similar data name classifying device of the present invention constructs a given data name and a dictionary in which terms constituting the data name and the type of the term are registered. A means for determining the type of a term using the dictionary, a means for extracting only a highly important term using the determined type of the term, and a similar method based on the extracted highly important term And a means for classifying the data names.

[0007]

In the present invention, a term that represents a value such as a number or an address that appears in a data name (referred to as a classifier) and a term that represents a subject that represents a data name (referred to as a main term) are identified. Classify data names with emphasis on the term. That is, when a data name and terms that make up the data name are given as input, a dictionary is used to extract a classifier and a main term, and then these two extracted terms are respectively extracted. The data names are converted into more general terms and the data names are classified based on the two terms resulting from the conversion.
This enables highly accurate classification of similar data names.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a similar data name classification device of the present invention. In the figure, 1 is a classification device main body, 2 is an input device, 3 is an output device, 4 is a dictionary, and the classification main body 1 is composed of a dictionary check unit 11, an important term extraction unit 12, and a classification processing unit 13.

FIG. 2 shows the structure of one entry in the dictionary 4. The term number 41 showing the number for uniquely identifying the term, the term name 42 showing the name of the term, and the term can be a section word. A segment word flag 43 (○ is acceptable, ×
Point), the main word flag 44 (○ is acceptable, x is not) indicating whether or not the term can be a main word, and the term number of a more generalized term (referred to as a representative term) It consists of a representative term number 45. The term for which the section word, that is, the section word flag 43 is "○", is "number",
It is a term such as “number” and “name” that comes to the right end of the data name and indicates what value the data name represents. Further, the main word, that is, the term in which the main word flag 44 is “◯”, is to the left of the segment word in the data name, and represents what the data is, such as “customer” or “order”. Is a term.

Next, the dictionary check unit 11, the important term extraction unit 12, and the classification processing unit 13 of the classification apparatus body 1 in FIG.
The operation of will be described in detail.

When the dictionary checking unit 11 receives the name of data, the terms forming the name of the data, and the arrangement order of the terms from the analysis designer of the database through the input device 2, the dictionary checking unit 11 determines each of the terms forming the data name. , The term that matches the term name 42 in the dictionary 4 is identified, and the terminology flag 43, the main word flag 44, and the representative term number 4 that the term has
A term name 42 (called a representative term name) of a term having the same value as 5 in the term number 41 is output. Here, the representative term number 4
When 5 indicates the own term number 41, the own term name 42 is the representative term name, and when the representative term number 45 indicates the term number 41 of another term, the term name 42 is the representative term name. Becomes

The important term extraction unit 12 sequentially checks the terminator flag 43 of each term from the right end of the arrangement of terms in the data name to the left, and extracts one term that can be the first term found. To do. Next, the terms to the left of the found term are sequentially examined toward the left, and one term that can be the first found main word is extracted. The important term extraction unit 12 further extracts a term name having the same term number as the representative term number of the term that can be the extracted term (this is called a representative term). Finally, a term name having the same term number as the representative term number of the term that can be the extracted main word (this is called a representative main word) is extracted.

In the classification processing section 4, the data names are subjected to sort processing in which the representative divisional words are classified into large classifications and the representative main words are classified into small classifications, and the results are output to the database analysis designer via the output device 3. The analysis designer determines that the data names having the same representative section word and the representative main word are similar.

FIG. 3 shows a specific example of the data name to be processed and the terms constituting the data name, and FIG. 4 shows a specific example of the contents of the dictionary 4. Hereinafter, examples of processing will be described based on these.

It is assumed that the data analysis designer of the database inputs the data names and the terms constituting the data names shown in FIG. 3 through the input device 2. Dictionary checker 11
Then, the terms that make up the data name shown in FIG. 3 and FIG.
The terminology of each term, the main word flag, and the representative term name are checked by matching the terms in the dictionary 4 shown in FIG. The result of the examination is as shown in FIG. For example, the data name "new application customer No." is composed of four terms "new", "application", "customer", and "No." It can be seen that “application” and “customer” are terms that can only be the main terms, and “No.” is a term that can be a segmental term and a main term.

The important term extraction unit 12 searches for a term that can be a section word and a term that can be a main word based on the information in FIG. 5, and further, based on that, a representative section word and a representative main word. Ask for. The results obtained are shown in FIG. For example, in the case of the data name “new application customer No.”, the term that can be the first term to be found by checking from the right end of the data name from FIG. 5 is “No.”, and its representative term is “No. It is. In addition, the term that can be the first major word found by searching from the left of this "No." to the left is "customer".
And its representative main word is also “customer”.
Although “application” is a term that can be a main word, it is not a main word in this data name.

In the classification processing unit 13, based on the information shown in FIG.
The input data names are sorted with the representative demarcation word as the major classification and the representative main word as the minor classification. FIG. 7 shows the sorted output result. FIG. 7 shows that “current customer number” and “new customer No.” are similar to each other, “reception number” and “contract No.” are similar to each other, and there is no data name similar to “customer address”. ing. The sorting result is shown to the database analysis designer through the output device 3.

[0019]

As described above, according to the similar data name classifying apparatus of the present invention, it is possible to classify data names from the viewpoint of the similarity of data values, and the database analysis designer can The data names can be unified efficiently.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of an embodiment of a similar data name classification device of the present invention.

FIG. 2 is a diagram showing a structure of a dictionary.

FIG. 3 is a specific example of data names to be processed and terms constituting the data names.

FIG. 4 is a specific example of a dictionary.

FIG. 5 is an output example of a dictionary check unit.

FIG. 6 is an output example of an important term extraction unit.

FIG. 7 is an output example of a classification processing unit.

[Explanation of symbols]

 1 classification device main body 2 input device 3 output device 4 dictionary 11 dictionary check unit 12 important term extraction unit 13 classification processing unit

Claims (1)

[Claims] 1. A dictionary in which a term that constitutes a data name and a term type are registered, a means that determines the type of a term that constitutes a given data name using the dictionary, and a means for determining the term. A similar data name classifying device comprising: means for extracting only terms of high importance using a type; and means for classifying similar data names based on the extracted terms of high importance.