JPH0799521B2 - Similar character string search device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] The present invention is a similar character string search device that searches for registered items similar to a keyword cut out from an input character string, when a desired item is selected from a database. However, if a desired one is selected from a menu provided over a plurality of pages by adopting a menu selection method, there is a problem that the selection becomes extremely complicated,
Moreover, in order to solve the problem that the system for language recognition becomes large and complicated if the input character string is recognized and the desired item in the database is selected,
With respect to the keyword cut out from the input character string, the most similar item among the items of a predetermined level lower than the item of interest provided hierarchically in the database is detected, and then this item is focused and repeated. By searching for a desired item with the keyword, a similar desired item associated with the data structure of the database is searched for the keyword cut out from the input character string.

[Industrial field of application] The present invention enables a desired item to be searched for by detecting the most similar item from the items hierarchically provided in a database for a keyword cut out from an input character string. The present invention relates to a configured similar character string search device.

[Problems to be solved by conventional technology and invention]

Conventionally, in order to read the data registered in the database, or to add or modify it, it is necessary to specify the target item, and there are a menu selection method, a natural language input method, and the like.

In the menu selection method, all the items attached to the data registered in the database are displayed, and a desired item in a desired page is designated by key input from the displayed items. With this method, the number of items to be displayed on the menu screen becomes large, and it is necessary to sequentially key in and select desired items over a plurality of pages, which makes the operation for selecting items extremely complicated. There was a problem that became.

In addition, the natural language input method has a problem in that it is necessary to recognize the input character string in a language, the system becomes large, and the CPU load for morphological analysis becomes large.

[Means for solving problems]

In order to solve the above problems, the present invention sequentially selects all items from a focused item in a database to a hierarchy lower than a predetermined level as candidate items corresponding to a keyword cut out from an input character string. Readout unit 4
An item read by the reading unit 4, a similarity calculation unit 5 for calculating the similarity with the keyword, and an item specifying unit for specifying the item based on the similarity calculated by the similarity calculation unit 5. 6 and the item specifying unit 6
Pay attention to the item specified by, read all items up to a predetermined level lower hierarchy, calculate the degree of similarity with the next keyword and specify the item repeatedly, and search for the desired item. I am trying.

FIG. 1 shows the principle configuration of the present invention. Key disassembly unit 1 in the figure
Is an input sentence such as "Why did you develop the plant for the demo of ESHELL?" From keywords such as "ESHELL", "Demo",
"Plant" and "development" are cut out.

The command determination unit 2 determines a command (search, registration, update, deletion, etc.) from the keyword cut out from the input sentence by the key decomposition unit 1.

The item search unit 3 searches the data 7 for an item corresponding to a keyword, and is composed of 4 to 6.

The reading unit 4 sequentially reads all items registered in the hierarchy from the focused item in the data 7 to a predetermined lower level.

The similarity calculating unit 5 calculates the similarity of each item read from the data 7 by the reading unit 4 for the keyword, and calculates the maximum similarity and the character string (character string of the item) at this time in the similarity table 8. To be stored in.

The item specifying unit 6 specifies an item corresponding to the keyword based on the maximum similarity calculated by the similarity calculating unit 5 and the maximum similar character string at this time.

The data 7 is items and the like stored in the database.

The similarity table 8 stores the maximum similarity calculated by the similarity calculator 5 and the maximum similar character string at this time.

[Action]

 Next, the operation will be described.

In FIG. 1, an input sentence, for example, "Have you developed the plant for the demo of ESHELL?" Is input, and the key decomposition unit 1 decomposes into keywords, such as "development", "plant", "demo" The command is decomposed into "ESHELL" and notified to the command determination unit 2. The command determination unit 2 that has received the notification of these keywords determines the command, for example, based on the "development" at the beginning of the keyword, with respect to the input sentence. The process to be executed is determined to be a search (search command) and is notified to the item search unit 3. The reading unit 4 configuring the item search unit 3 detects all candidate items corresponding to the keyword, for example, the item of interest to the second lower level. Items are sequentially read out from the data 7 and notified to the similarity calculation unit 5. The similarity calculation unit 5 calculates the similarity with the keyword for each of the notified items, and calculates the maximum similarity value and the maximum similarity value. The character string of the item at this time is stored in the similarity table 8. The item specifying unit 6 specifies the noteworthy item based on the maximum similarity stored in the similarity table 8 and the maximum similar character string at this time. Similarly, all the items from the item of interest to the lower two levels are read for the next keyword, and the maximum similarity and the maximum similar character string at this time are stored in the similarity table 8.
And specify the corresponding item.

As described above, with respect to the keywords cut out from the input sentence, all the items from the item of interest to the hierarchy lower by a predetermined level are sequentially read out, the most similar item is specified, and then the specified item is focused. Similarly, by repeatedly specifying the most similar item, it becomes possible to easily search the item most similar to the keyword cut out from the input sentence using a simple system.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be sequentially described in detail with reference to FIGS. 2 to 10.

FIG. 2 is a conceptual explanatory view of the present invention. This shows the operation and display when the input sentence "Is there a search tool for reletional DB?" Is input from the keyboard as shown in (a) Input sentence on the display 9.

The “decompose to key” in the figure shows a state in which “relational DB” and “search tool” are cut out from the input sentence in (a) and decomposed into keywords.

Matching in the figure is the most similar item "tool" corresponding to the "search tool" of the decomposed keyword in the figure
Indicates the state of identifying. Of the items registered in the data structure of the database, as described later,
This means selecting the one with the highest degree of similarity.

Matching in the figure shows a state in which the most similar item “RDB” (this is registered as an alias “relational DB”) corresponding to the next keyword “relational DB” decomposed in the figure is selected.

By the above matching, the matching with the keyword cut out from the input sentence is completed, so this item "RDB"
Information registered in association with (alias “relational DB”) is displayed on the display 9 as shown in FIG. The user corresponds to the information "Tools related to RDB include the following" menu shown in (b) display (1) in this figure, for example, the number "1"
Key in. As a result, information such as the information “X is a tool for doing” shown in (c) display (2) in the figure can be searched.

As described above, even if the input sentence is not accurately input, items most similar to the keywords cut out from the input sentence can be sequentially specified, and desired information can be searched and displayed.

Next, the operation of the configuration shown in FIG. 1 will be described in detail with reference to FIG.

In FIG. 3, the figure shows a state in which an input sentence is input.

The figure shows a state of being decomposed into keywords. this is,
Input sentence input to the key decomposition unit 1 in the figure. For example, the input sentence "Why did you develop the plant for the demonstration of ESHELL?" Shown in Fig. 4 (a) is shown in Fig. 4 (a) keyword. "Development", "Plant Part", "Demo", and "ESHEL"
It means to decompose into a key list consisting of L ".
Note that this decomposition is in reverse order to the input sentence. Similarly, the input sentence is decomposed into keywords as shown in FIG.

The figure shows the state in which a command is determined. This means that when the head of the key list decomposed in the figure is registration, update, or deletion (and its alias), the corresponding command is executed and the head of the key list is deleted. For example, since the head of the key list in FIG. 5 (a) is "registration", the registration is executed and the "registration" is deleted from the key list as shown in FIG. 5 (b). On the other hand, in other cases, for example, when the beginning of the keyword in FIG. 4 (a) is "development", the search is executed as shown in FIG. 4 (b). At this time, the top of the key list is not deleted.

The figure shows a state in which initial setting is performed. This shows a state in which the maximum similarity “0.0” and the maximum similar character string “” (this means an empty character string) in the similarity table 8 in FIG. 1 are initialized.

The figure shows a state in which it is determined whether or not the key list to be selected is empty. This is the key list that decomposes the input sentence and deletes it in the case of registration, for example, for key lists 10-1 and 11-1 shown in FIGS. 4 (c) and 5 (c),
This means that all the similarities are judged, and it is judged whether or not they are deleted and become empty. If YES, end (END). If NO, execute in the figure.

The figure shows a state in which the degree of similarity is calculated. This similarity is determined by the keyword “ESHE” in the key list 10-1 in FIG.
As indicated by the arrow from LL ", the item of interest, for example, all items registered in the hierarchy from" excellent "to the lower two levels (except the item of interest has already been specified, the same applies below). On the other hand, it means calculating the degree of similarity (which will be described later with reference to FIGS. 6 to 8).

The figure shows a state in which it is determined whether or not the similarity is higher than the maximum similarity. This means that it is determined whether or not the similarity calculated in the drawing is larger than the maximum similarity stored in the FIG. 1 similarity table 8. YES
In the case of, the similarity calculated in the drawing is set to the maximum similarity in the FIG. 1 similarity table 8, the item at that time is set to the maximum similar character string, and the following steps in the drawing are repeated. If NO, repeat the following in the figure.

By the above procedure, the maximum similarity that is most similar to the keyword and the maximum similar character string (item) having this maximum similarity are stored in the similarity table 8 in FIG. More specifically, for example, for the first keyword “ESHELL” in the key list 10-1 in FIG. Of “ES
The half-width "ESHELL" registered as an alias for the key (item) "HELL" is the most similar (match). As shown by the arrow to the second keyword "demo", the item of interest " The key (item) "DemoPro" one level below ESHELL is the most similar. As shown by the arrow for the third keyword "Plant No.", one level below the item "DemoPro" to be noted. Is most similar to the key (item) of “Petroleum plant production plan.” Further, as shown in FIG. Items such as "developer, development man-hours" included in the information of "oil plant production plan" are most similar.

Further, in the case of FIG. 5C, the registration command is extracted in FIG. 5B and it is determined that the registration is to be executed. Plant fault diagnosis ". Then keylist 11
-For the first keyword "ESHELL" in -1, as indicated by the arrow, a half-width character that is registered as an alias for the key (item) named "ESHELL" two levels below the item of interest "excellent""ESHELL" is the most similar (match). As a result, there are no more keywords to match in the key list 11-1 (it became empty), so as shown in Fig. 5 (d), the item "DemoPro" shown below this "ESHELL" Register the item "plant fault diagnosis" below and connect it with a path. Further, the target data for this item "plant fault diagnosis" is also connected by a path as shown in the figure.

The data search by the key will be specifically described with reference to FIGS. 6 to 8.

In Fig. 6, "FOCUS" is the item of interest, SUBI is the item one level below the "FOCUS" of interest, and SSUBI.
Indicates an item in a layer two levels below the focused "FOCUS". NEWFOCUS is an item of interest for newly specifying FOCUS and performing the next data search based on the result of obtaining the item with the highest degree of similarity. The black circles "●" in the figure are the matched items. This will be described below using the search flow of FIG.

In FIG. 7, “SUBI, SSUBI described in (1) in the figure
When there is an item that matches the key together, and "SUBI
Item has a higher degree of similarity with the key than the item of SSUBI ”, and“ If the item of SSUBI is lower than the item of SUBI ”,“ SUBI is confirmed and SSUBI is processed as a new FOCUS. Continue ”(Fig. 6 (a)). Fig. 6 (a) shows that the black circles match, and the black circles in SSUBI are identified as new attention points (NEWFOCUS).

Similarly, by applying the logic shown in FIG. 7, the item described as NEWFOCUS in FIGS. 6B to 6E on the right side is specified as a new item of interest.

As described above, the items specified using the search flow shown in FIG. 7 are sequentially searched as new items of interest, so that similar items can be easily found from the associated and registered data. Is possible.

FIG. 8 shows a specific example of the similarity calculation between the “fault diagnosis” and the “stacker crane fault diagnosis”.

In the figure, (a) shows a state in which the "fault diagnosis" for calculating the similarity and the "stacker crane fault diagnosis" are input to the similarity calculation unit 5 in FIG.

In the figure, (b) shows a state in which evaluation points are calculated. this is,
This means that the evaluation score is calculated using the following formula.

Evaluation point = 1 / (number of characters in short character string) (1) In this example, since the short character string “fault diagnosis” has 5 characters, this value “5” is added to the expression (1). Substitute and evaluate points = 0.
2 is obtained.

In the figure, (c) shows a state of initial setting. This is the first
This means that the maximum similarity in the figure similarity table 8 is set to 0.0.

In the figure, (d) shows a state in which the “fault” matches and the degree of similarity is calculated. This similarity (n) is calculated using the following formula.

Similarity (n) = similarity (n−1) + partial similarity− (similarity (n−1) × partial similarity) (2) In this example, similarity (1) = 0.0 + 0.2 × 2-0.0 × 0.2 × 2 = 0.4 Here, the partial similarity substituted into the equation (2) is a partial character string that matches the evaluation score “0.2” calculated using the equation (1). It is the length of the "fault" multiplied by "2".

In the figure, (e) shows a state in which the next character string “diagnosis” matches and the degree of similarity is calculated using equation (2).

Similarity (2) = 0.4 + 0.2 × 2-0.4 × 0.2 × 2 = 0.64 In the figure (f), the similarity is calculated in consideration of the difference in the length (the number of characters) of the two character strings. Show. Here, the formula (2)
The partial similarity substituted for is a constant, for example "0.3", divided by the difference (12-5 + 1) in the length of two character strings.

Similarity (3) = 0.64 + 0.3 / (12-5 + 1) −0.64 × 0.3 /
(12−5 + 1) = 0.65 By the above procedure, the similarity (3) between the two is calculated as 0.65, and if the threshold value is set to, for example, 0.6, it is regarded as a match.

Next, an execution example will be described with reference to FIGS. 9 and 10.

In Fig. 9, when the input sentence "Which function handles a file?" Shown in Fig. 9A is keyed in, "File" and "Function" are cut out as keywords, and the data structure is traced (b) ) Display using (1) FG: COPYFIL
The data "E" is reached and displayed on the display as shown in the figure. Furthermore, the contents are displayed in (c) in the figure (2).
It is displayed as shown in. In this way, the most similar item is displayed from the input sentence and the content thereof is displayed, so that character input when searching the database becomes extremely easy.

FIG. 10 (a) shows the most similar item "input / output processing" to "output processing" (NEAREST), which is calculated using the procedure described above. . Of the list processing, character string processing, input / output processing, editing processing, program management, data type conversion, and determination items in the figure, the most similar item to "output processing" is "input." Output processing ”, and the similarity is 0.68.

FIG. 10 (B) shows that the degree of similarity that "output processing" is similar to "list processing" is 0.19. This is smaller than the degree of similarity shown in FIG.

Similarly, in Fig. 10 (c), the similarity between "input / output processing" and "input / output disposal" is 0.70, and the similarity between "input processing" and "input disposal" is 0.41. It means that.

Of the above, if the threshold value is set to, for example, 0.6, it is determined that “output processing” matches (input / output processing) and “input / output processing” matches “input / output disposal”. Item is displayed on the display, and the contents registered in association with each other are displayed.

〔The invention's effect〕

As described above, according to the present invention, a keyword cut out from an input character string is most similar to a keyword lower than a focused item among items hierarchically provided in a database. Since a configuration is adopted in which the specified item is specified and then the specified item is focused and sequentially searched, similar items related to the data structure of the database are searched for the keywords cut out from the input character string. It can be easily searched using a simple system.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a principle configuration diagram of the present invention, FIG. 2 is a conceptual explanatory diagram of the present invention, FIG. 3 is an operation explanatory flowchart of the present invention, and FIG.
FIGS. 5 and 5 are explanatory diagrams of the operation of the present invention, FIG. 6 is an explanatory diagram of data retrieval by keys, FIG. 7 is a data retrieval flow by keys, FIG. 8 is a similarity calculation example, FIGS. 9 and 10. Shows an execution example. In the figure, 1 is a key decomposing unit, 2 is a command determining unit, 3 is an item searching unit, 4 is a reading unit, 5 is a similarity calculating unit, 6 is an item specifying unit, 7 is data, 8 is a similarity table, 9 is a similarity table. The displays, 10-1, 10-2, and 11-1 represent key lists.

Claims (1)

[Claims] 1. A keyword extraction unit for extracting a keyword from an input character string, and sequentially reading all items from a focused item in a database having a hierarchical structure up to a predetermined lower level of a plurality of levels. A reading unit (4), an item read by the reading unit (4), a similarity calculation unit (5) for calculating the similarity with the keyword, and a similarity calculated by the similarity calculation unit (5). A similar character string search device comprising: an item specifying unit (6) for specifying an item based on