JP3331692B2 - Heterogeneous dictionary creation support system - 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 supporting creation of an allotted word dictionary used in an information search system by keyword search.

[0002]

2. Description of the Related Art An information retrieval system based on keyword retrieval is configured so that keywords and information (documents / documents) are stored in association with each other and stored information can be retrieved from the keywords. A keyword index covers a correspondence between a keyword and information corresponding thereto, that is, a correspondence between a keyword and a list of numbers of documents and documents having the keyword therein. In such a search system, stored information can be extracted only when the input keyword of the searcher completely matches the keyword (keyword in the keyword index) possessed by the system. If the notation fluctuates, a search intended by the searcher may not be performed. For example, if a document is associated with the keyword “simulation” in the information search system, the document cannot be searched if the input keyword is “simulation”. As a system that can cope with the fluctuation of the notation due to such a different notation word, there is an information retrieval system including a synonym dictionary. In a system provided with a synonym dictionary, a heterologous keyword is registered in advance as a synonym of the original keyword, and a search is performed by using the synonym keyword together with the search. In the above example,
Register "Simulation" as a synonym keyword for "Simulation"
In response to the input “Simulation”, a search is also performed for “Simulation”.

A synonym dictionary is generally created by manually registering the correspondence between a headword and a synonym (here, a different notation).

[0004]

According to the conventional method of manually creating a synonym dictionary (heterogram dictionary as a kind of dictionary), what words are used as headwords, and what are the corresponding keywords? It was up to the creator to do it. Therefore, it is very difficult to register all the notations. In addition, the allograph word keyword corresponding to the headword of the allograph word dictionary should be a keyword existing in the keyword index of the information search system. This is because documents corresponding to keywords that do not exist in the keyword index do not exist in the system, and there is no point in registering them. It is difficult to guarantee this with the manual registration method. There is also a support system for generating a notation word. Such a support system generates a variant word using a character conversion rule for a headword entered manually. However, such a notation word creation support system based on the character conversion rules is created for editor-like use, and is not used for keyword search, but is included in the keyword index because the notation words are synthesized independently. There is no guarantee.

[0005] An object of the present invention is to realize a high-quality heterographic word dictionary creation support device that generates heterographic words by determining the similarity between keywords in a keyword index.

[0006]

FIG. 1 is a block diagram of the present invention. The keyword index A is a file in which keywords for searching are collected in the information search system. Further, the heterologous word dictionary D output from the present apparatus is to be used as a synonym dictionary in the information retrieval system.

A first invention: a headword extraction means 1 for sequentially extracting keywords from a keyword index A of an information search system to obtain a headword B;
And the respective keywords in the keyword index A to calculate the similarity, and extract the notation word 2 having the similarity not less than a predetermined value as the notation word C; Dictionary registration means 3 for registering the different notation words C in correspondence with each other.

Second invention: In the first invention, the variant notation extraction means 2 sets the variant notation C when the attribute of the keyword extracted as having high similarity satisfies a prescribed condition.

Third invention: In the first invention or the second invention, an allotted word candidate presenting means 4 for displaying the allotted word extracted by the allotted word extracting means 2 as a candidate word on a display screen.
And a notation word selecting means 5 for leaving, as a notation word C, a candidate word selected by the operator among the candidate words displayed on the display screen.

Fourth invention: In the first to third inventions, an allotted word candidate presentation unit 4 for displaying the allotted word extracted by the allotted word extraction unit 2 as a candidate word on a display screen; And a notation adding means 6 for adding a new keyword as a notation C.

Fifth invention: In the fourth invention, the pseudonym adding means 6 checks whether or not the newly input keyword exists in the keyword index A. Is a headword, and the first headword and other candidate words are heterologous words.

[0012]

The headword extracting means 1 fetches the keywords in the keyword index A one by one and sets them as headwords. The notation extraction unit 2 extracts a keyword having a high degree of similarity from the keyword index. The dictionary registration unit 3 registers a keyword having a high degree of similarity as a notation word in association with the headword. If there is no unusual word, no headword is registered. When all of the keyword indices are processed in this way, an allotted word dictionary D is automatically generated.

According to a second aspect of the present invention, the quality of the spelling dictionary D is increased by excluding keywords having a high degree of similarity and having poor attribute conditions. The third aspect of the present invention enhances the quality of the spelling dictionary D by displaying the automatically extracted similar words on the screen of the display device so that the judgment of the operator can be entered.

According to a fourth aspect of the present invention, an operator can add an alternate notation word, thereby providing an alternate notation dictionary D
To increase the quality. According to the fifth invention, when a keyword added as an allotted word by the operator does not exist in the keyword index A, the quality of the allotted word dictionary D is increased by using the keyword as a headword.

[0015] With this configuration, the generated pseudonym dictionary D can be used as an effective synonym dictionary in the information retrieval system. Since the similarity between keywords that originally exist in the keyword index A is determined to be a different notation, no useless words are included.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The processing procedure will be described with reference to the block diagram of FIG. Headword retrieval means 1 is a keyword index
Set the keyword extracted from A into the buffer.
This is headword B. The variant notation extracting means 2 compares the keywords (headwords) in the buffer with the keyword group of the keyword index A and extracts similar keywords.

Known means may be used to determine the similarity between keywords. FIG. 2 is an explanatory diagram of the calculation of the similarity. In order to determine the similarity between two character strings, characters are associated with each other, the similarity is evaluated, and the similarity is expressed by a numerical value (similarity calculation). LCS (Longest
Common Subsequence: The longest common substring method and the Heckel method. The similarity evaluation quantifies the similarity of a pair of character strings for which correspondence has been completed, using a similarity factor. Elements of the similarity include a matched character string length, an unmatched character string length, and the number of matching sections. The similarity is higher as the matching character string length is longer and the non-matching character string length is shorter. Even with the same matching character string length, the smaller the number of matching segments, the higher the similarity. For example, a value obtained by multiplying each similarity element by a weight coefficient and adding the result is set as the similarity. Even keywords extracted as similar keywords that are used less frequently or that have a large difference in the number of characters are not valid and are therefore deleted. FIG. 3 shows an example of determining an effective keyword. (A) is a similar keyword extracted for the headword "simulation", but "simulation" has only one appearance frequency. Also, the "configuration" has a string length that is too different. Such attribute conditions are set in advance, and only those that satisfy the conditions are left. In this example, if the attribute condition is set to an appearance frequency of 3 or more and a character string length difference of 2 or less, “simulation” and “emulation” shown in FIG. The alternative word candidate presentation means 4 displays the alternative word candidate on a display device. The alternative word selection means 5 deletes some of the candidate keywords according to the operator's instruction. The variant notation adding means 6 adds a candidate keyword according to an instruction of the operator. Further, it is checked whether or not the added candidate keyword exists in the keyword index D. If it does not exist, a warning is given, and if instructed, this word is replaced with the headword. The dictionary registration means 3 associates the headwords determined as described above with the different notation words to make entries in a different notation word dictionary and registers them in the different notation word dictionary D. FIG. 4 shows an example of an entry of the heterographic word dictionary. By performing the above processing for all the keywords in the keyword index D, an allotted word dictionary is completed.

FIG. 5 shows an example of a screen display in the above steps in the above processing. (A) is a state in which the alternative notation word candidate presenting means 4 displays the alternative notation word candidate on a display device. The keyword "simulation" is extracted from the keyword buffer and displayed as a heading. For the headword "simulation", "simulation", "simulation", and "emulation" are displayed as alternative notation word candidates. All the notation words displayed on the screen are in a selected state (default setting). If this is the case, the operator issues a registration instruction by hitting the return key or the like. If any of the displayed keywords are inappropriate,
After inputting a non-selection instruction, a registration instruction is performed. Here, when the operator does not want to register "emulation" as a variant word of "simulation" and instructs non-selection, inappropriate keywords are deleted as shown in (B). (C) shows a case where the operator comes up with “simulation” as a different notation and newly registers it.
The new keyword entered is shown on the input line of the screen. If this input keyword exists in the keyword index, it becomes a registration target.

If it does not exist, a warning is issued on the screen, and if there is an instruction from the operator, it is replaced with the currently displayed headword. (D) shows the state after the instruction. Although the embodiment to which all of the invention is applied has been described, it goes without saying that only a part may be applied. Further, in the present embodiment, the apparatus has been described as an independent heterographic word dictionary creation support apparatus, but it should normally be incorporated as a part of the information search system. In that case, the heterologous word dictionary forms a part of the synonym dictionary.

[0020]

As described above, according to the present invention, it is possible to create an allotted word dictionary that comprehensively includes allotted words. Moreover, since the non-words always exist in the keyword index, they do not include useless keywords. Therefore, a high-quality notation dictionary suitable for keyword search can be created.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of the present invention.

FIG. 2 is an explanatory diagram of similarity calculation.

FIG. 3 is an explanatory diagram of valid keyword determination.

FIG. 4 is a diagram showing an example of a notation word dictionary entry.

FIG. 5 is a diagram showing an example of a screen display for explaining the operation of the embodiment.

[Explanation of symbols]

DESCRIPTION OF REFERENCE NUMERALS 1 headword extracting means 2 allotted word extracting means 3 dictionary registration means 4 allotted word candidate presenting means 5 allotted word selecting means 6 allotted word adding means A keyword index B headword C different notation word D different notation dictionary

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 17/30

Claims (5)

(57) [Claims] 1. A keyword is sequentially extracted from a keyword index (A) of an information search system and a headword is extracted.
(B), the headword extraction means (1), the headword (B), and each keyword in the keyword index (A) are compared to calculate similarity, and the similarity is equal to or greater than a predetermined value. A different notation word extracting means (2) for extracting a word as a different notation word (C); and a dictionary registering means (3) for registering a headword (B) and a different notation word (C) in association with each other. Notation dictionary creation support device. 2. An allotted word extracting means (2), wherein an attribute of a keyword extracted as having a high similarity satisfies a prescribed condition, is set as an allotted word (C).
2. A device for supporting the creation of a dictionary of different notations described in 1. 3. An allotted word candidate presenting means (4) for displaying the allotted word extracted by the allotted word extracting means (2) as a candidate word on a display screen, and among the candidate words displayed on the display screen, 3. An apparatus for supporting the creation of an allotted word dictionary according to claim 1 or 2, further comprising an allotted word selecting means (5) for leaving an item selected by the operator as an allotted word (C). . 4. An allotted word candidate presenting means (4) for displaying the allotted word extracted by the allotted word extracting means (2) as a candidate word on a display screen; 4. The apparatus according to claim 1, further comprising: a means (6) for adding a notation word added as C). 5. An allotted word adding means (6) checks whether or not the newly input keyword exists in the keyword index (A). If not, the input keyword is used as a headword, 5. The apparatus according to claim 4, wherein the first headword and the other candidate words are different notations.