JPH0785092A - Method and device for retrieving ambiguous graphic - Google Patents

Abstract

PURPOSE:To provide the method for retrieving ambiguous graphic capable of retrieving graphics corresponding to the ambiguous words inputted by a user from among an electronic document which is made by the texts and graphics. CONSTITUTION:The electronic document which is made by the texts and graphics is stored in a multimedia document DB185. With a keyboard 101 and a pointing device 102, retrieval words to retrieve the required graphics are inputted. A retrieval condition preparing part 131 edits the retrieval conditions specifying the location between the inputted retrieval words and the required graphic. A document retrieval part 185 retrieves the text in the electronic document including the retrieval words and identifies the graphic corresponding to the retrieval conditions edited by a graphic collation part 152. A display device 103 displays the graphic identified.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document management system for collectively storing electronic documents, and more particularly, to search for drawings, tables, photographs, etc. in electronic documents stored in a computer system to provide a multi-window function. TECHNICAL FIELD The present invention relates to a fuzzy figure search method and apparatus for freely displaying search results.

[0002]

2. Description of the Related Art Japanese Unexamined Patent Publication No. 2-214979 discloses a technique for retrieving graphics stored in an electronic document. This is to store a keyword and image information such as a figure in pairs, and to search the image information using the keyword. However, in this method, since there is no correspondence between the keyword and the image information, it may be difficult to remember the keyword when searching, and it may be difficult to find desired image information.

In addition, the user draws an outline of a desired figure,
"Image retrieval using schematic images" is used for retrieval based on this (Kyoji Hirata et al., IEICE Tech. Vol.92, No.51, DE92-2,
There is an image retrieval system described in 1992.5). However, in this system, the user does not have a drawing heart and cannot draw the outline of the figure well, so the system presents a figure different from the figure intended by the user as a search result. There was a problem with.

[0004]

As described above, in the conventional technique, it is possible to search for a graphic in a specific field by associating a word with a graphic, but it is troublesome to associate the graphic.
There was a problem such as lack of versatility of search.

The object of the present invention is to solve the above-mentioned problems as much as possible, to save the labor required for associating a word with a graphic, to limit the field, and to search for a desired graphic even from ambiguous words. A method and apparatus are provided.

[0006]

An electronic document composed of text and graphics is stored in advance, and at least one search word for searching a predetermined electronic document from the stored electronic document is input and input. The search condition that specifies the positional relationship between the search word and the desired figure is interactively edited, the electronic document including the input search word is searched, and the figure corresponding to the edited search condition is searched for. The feature is that the figure is identified from the figures of the electronic document and the identified figure is displayed. With this configuration, the above object is achieved.

[0007]

The operation and principle of the fuzzy figure retrieval method of the present invention will be described below. The keyword dictionary 170 is a systematization of words around us based on their meanings. For example, the word "cat" is synonymous with "cat" and "ca.
t ”etc. The word “mouse” has synonyms “mouse”, “rat”, “mouth”, “mouse” and the like. However, the mouse as an "animal", the "body"
The word "mouse" can be classified into a plurality of meanings, such as a mouse as a mouse and a mouse as a "pointing device". An abstraction of a certain word, such as “animal”, “body”, and “pointing device”, is called a broader term. In such a case, when the above synonyms are classified in correspondence with the broader terms, the mouse synonyms as “animal” are the mouse synonyms as “mouse”, “rat”, “mouse”, and “body”. Is synonymous with mouse as "mouth", "mouth", "pointing device"
"se". Also, for example, the subordinate word of the mouse as “animal” is “mus musculus”, and the subordinate word of the mouse as “pointing device” is “3 button mouse”.
Like, the subordinate word also differs depending on the meaning of the word. In this way, synonyms and subordinate words corresponding to the meanings of the words are stored. In addition, the word "hanten" is a homonym "reverse",
It has "spots" and these are also stored as different ones. In this way, the keyword dictionary stores synonyms, broader terms, lower terms and homonyms of words.

FIG. 3 shows the flow of processing for retrieving a figure, together with the relationship between the keyword dictionary, the figure correspondence table and the allocation information. For example, when searching for a figure related to “mouse” and “hunting”, the search word acquisition 310 acquires the character strings “mouse” and “hunting” input by the user. In the next different notation unifying process 320, it is determined whether or not there is a change in the notation such as "va" and "ba" in the character string input by the user, and if there is a change, refer to the keyword dictionary. However, the keywords are unified according to the notation set by default. Hereinafter, in the present specification, a word input by the user will be referred to as a search word, and a word after performing the different notation unification process with reference to the keyword dictionary will be referred to as a keyword to be distinguished. In this case, since there is no fluctuation in the notation, it is not necessary to perform different notation unification processing.

In the next keyword expansion processing 330, the presence or absence of homonyms in each search word is checked. in this case,
"Mouse" has three meanings: a mouse as an animal, a mouse as a body, and a mouse as a pointing device. Therefore, a list of high-ranking words of "mouse" is displayed, and the user can select from among them, so that the meaning of "mouse" can be restricted. In addition, "hanten" has the synonyms of "reversal" and "speckle." Therefore, similar to the case of "mouse", a list of homonyms related to "Henten" is displayed, and the user can select from them to restrict the meaning of "Hunten". In this way, synonyms corresponding to the meaning of each search word are expanded by adding restrictions to the meaning as needed. Further, if requested by the user, a subordinate word of each search term can be added as a keyword. For example, when adding a subordinate word of mouse as an animal, a list of subordinate words of the search word such as "mus musculus" and "guinea pig" is displayed, and the subordinate word of "mouse" is selected by the user from the list. Can be added to keywords. Next, in the text search processing 340, the keywords in the documents stored in the (multimedia) document DB are searched based on the keywords thus developed. For specific text search processing method, for example,
JP-A-3-15980 or Japanese Patent Application No. 2-1100
The method described in No. 10 can be adopted.

Next, in the narrowing-down process 350, the graphic information in the document searched by the text search process is acquired,
Of the graphic directives that designate a graphic such as "Fig. 1" and "Photo 2", the layout information 180 is referenced to search for graphic directives near the keyword, and the graphic correspondence table is referenced to search for the graphic. Identify the graphics that match the directive. The distance between the keyword and the figure designator is given by the number of lines, the number of pages, the number of characters, etc. Unless otherwise specified, the figure designator is searched using the default value as the neighborhood condition. In addition, if the search condition for the graphic is given, that is, if at least one of the type of the graphic, the distance between the keyword and the graphic, or the position of the graphic on the page is given as the search condition, the content of the search condition Further narrow down according to. First, when the type of figure, for example, "figure", is given as a search condition, the figure-directed word to be searched is only the figure-directed word representing "figure", and the figure-directed words near the keyword are specified. Search for. Second, the distance between the keyword and the figure,
For example, when "within one page" is given as the search condition, the layout information is referred to, the distance between the keyword having the graphic directive in the vicinity and the graphic corresponding to the graphic directive is measured, and the page Both positions are determined, and only the figures whose distance or positional relationship satisfies the search condition are narrowed down. Thirdly, if the position of the figure on the page, such as the upper part or the central part, is given as the search condition, refer to the allocation information,
Of the retrieved documents, the documents that include the graphic at the specified position are narrowed down, and the graphic designator is searched on the document. In this way, the figures are narrowed down.

Next, in the case number display 360, the number of retrieved figures is displayed at a predetermined position. Next, the searched figure display 37
At 0, a part of the retrieved figure is cut out to a predetermined size or enlarged / reduced to a predetermined size and displayed as a list. According to the user's request, the document name including the graphic, the application name that created the document, and the like can be displayed.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration of the apparatus and the outline of the operation of the system will be described below with reference to the drawings. FIG. 2 is a system configuration diagram of a document management system which is an embodiment of the present invention. The system includes a control device 201, a main storage device 205, a magnetic disk group 206, a display device 103, a keyboard 101, and a pointing device 102 such as a mouse. The system stores and stores documents in the multimedia document database 185. The keyword dictionary 170 accumulates and stores words centering on words used in the field of the user as keywords. Then, the user inputs a word related to the figure, the system expands the entered word or synonym as a keyword, interactively edits the search condition regarding the figure if the user requests, and stores the stored document. A document including a keyword is searched from among the figures, and a figure satisfying the search condition for the figure is searched from among the figures in the document. The multimedia document database, keyword dictionary, graphic correspondence table, and allocation information are stored in the magnetic disk group, and the contents of the keyword dictionary are read into the main memory. Text and graphics in a document registered in the multimedia document database are displayed, and a processing request is issued using a display device, a keyboard, and a mouse.
The search result is displayed on the display device. This system uses the present invention when searching for a figure in a document.

FIGS. 4 to 7 are views showing the internal data structure of the system, FIGS. 8 to 17 are display screen examples, and FIG.
8 to 23, the outline of the operation will be described using the PAD showing the flow of processing.

The main module of the system comprises an interactive display section 110, a graphic search section 150 and a document management section 160. The dialogue display unit 110 includes a dialogue control unit 120 and a search condition display control unit 130. The dialogue control unit controls the dialogue with the user, and includes an input analysis unit 121 and a window / menu control unit 122. The search condition display control unit displays and edits search conditions and displays search results, and includes a search condition creation unit 131 and a search result display unit 132. The graphic search unit is composed of a keyword expansion unit 151, a graphic collation unit 152, and a document search unit 155. A keyword dictionary, a graphic correspondence table, layout information, and a multimedia document database are stored in the magnetic disk group. The information stored in the keyword dictionary, the graphic correspondence table, and the allocation information is developed in the main storage device 205 by the graphic search unit and managed as a data structure according to the type of information.

A keyboard 101 or a pointing device 102 is used by a user to input a search word or edit search conditions. The input information is analyzed by the input analysis unit and passed to the search condition creation unit. The keyword expansion unit manages information on keywords and information on different expressions, unifies different expressions with respect to a search word input by the user, and expands to a lower word corresponding to the meaning of the word. Here, the keyword is passed to the document search unit, and a document having the keyword is searched from the documents managed by the document management unit. The document retrieved in this way is passed to the figure collating unit. The figure collation unit manages information about figure directives, correspondence information between document types and figure directives, document allocation information, correspondence information between figure directives and figures, and the like, depending on the contents of search conditions specified by the user. Then, the retrieved documents and figures are narrowed down. Here, we identify the graphics that meet the search conditions,
Generate a set of documents containing graphics. In the search result display area,
Display the searched figure.

In this embodiment, a method of narrowing down documents having a figure after performing a text search is adopted, but a method of performing a text search after narrowing down documents having a figure is adopted depending on the management method of documents. You can also

FIG. 8 is an interactive screen with the user when creating search conditions. The figure search window 800 includes a document type designation area 810, a search word input area 820, a figure type designation area 830, a range designation area 840, and a search result display area 85.
0, a function list display area 860 and the like. The user selects an item displayed in the document type designation area with a pointing device to set the document type to be searched. A mark such as a check mark is given to the selected item. Further, it is displayed so that it can be seen which item can be designated as the search condition according to the set document type. The user specifies a search word by inputting a character string in the search word input area. When inputting a plurality of search words in this area, there are methods such as enumerating words using specific symbols or expressing using natural language. The user selects the item indicating the type of the graphic displayed in the graphic type designation area 830 with the pointing device, thereby setting the type of the graphic to be searched. When narrowing down, it narrows down to the figure corresponding to the type of figure set here. In the range designation area 840, the distance between the keyword and the figure, that is, the positional relationship is set, and when narrowing down, the figure is narrowed down to the corresponding figure based on the distance set here. The range designation area includes a line range designation area 842 for designating a range by the number of rows and a page range designation area 845 for designating a range by the number of pages.

In the present embodiment, the method of designating the distance between the keyword and the figure by at least one of the number of lines and the number of pages is adopted. However, the document to be searched is a different document by using a hyperlink or the like. If the figure is linked to the figure, the method of making the figure linked to from another document the search target by displaying the item for specifying the necessity / non-necessity of the hyperlink search and selecting it by the user is adopted. You can also

Further, in this embodiment, the method of distinguishing the selected / non-selected state of an item by the presence or absence of a check mark is used. However, the display color of the button corresponding to the item and the background color are reversed and displayed, It is also possible to adopt a method in which the display color and the background color of the character string representing the content of the item are reversed and displayed.

In the search result display area 850, the document type specified in the document type specification area, the search word input in the search word input area, the figure type specified in the figure type specification area, and the range specification area are specified. The figure searched by using the specified range as the search condition and the number of cases are displayed. When displaying the searched graphic in the search result display area, the size is appropriately changed according to the size of the graphic, and then the list is displayed. For example, a part of the search figure is cut out and displayed in a predetermined size, or a figure obtained by enlarging / reducing the original figure (searched figure) is displayed. When displaying the searched figure or the document including the figure, select the figure to be displayed with the pointing device and the corresponding menu item in the function list display area 860 with the pointing device to display the same area or another screen. Can be displayed on. Display multiple figures that have been reduced,
A method for selecting a desired figure from among them is disclosed in JP-A-63-63
There is -4334 publication.

Next, as a first embodiment of the search method of the present invention, taking a case of searching for a figure related to "mouse" and "hanten" as an example, a method for editing search conditions, a search method and a search result. The display method of will be described.

First, in response to a user's request, a search condition creation screen, in this case, a graphic search window 800 is displayed in step 1800 (see FIG. 18), and the user waits for input. User input information is acquired, step 18
At 05, it is determined whether the input information is a session end request. If it is an end request, the process moves to step 1890, and if not, it moves to step 1810. In this case, since it is not a session end request, step 18
Go to 10. Next, according to the document type specified by the user,
In step 1910 (see FIG. 19), the document database storing the document of the type specified is identified. In step 1920, the graphic correspondence table is referenced to check the items that can be designated as the search condition for the document of the designated type according to the type of the document so that the user cannot select the items of the search condition that cannot be designated. After that, the input is awaited by the user. Next, the user input information is acquired, and step 1
At 930, it is determined whether the input information is a search execution processing request. If it is a search execution request, step 18
20 (see FIG. 18), otherwise step 1
Move to 940. In this example, since it is not a search execution processing request, the process moves to step 1940.

In this case, the user's input information is a search condition or a search word change request, and the type is determined.
In this example, the request is a change request for the search term, the type of figure, the distance to the figure, or the document type.

First, a change request for a search word is generated by selecting the search word input area 820 with a pointing device and inputting a character string from a keyboard or the like.
In this case, in step 1950, it is determined whether the method of specifying the search term is appropriate. If the designation method is not appropriate, an error message is displayed in step 1955. If so, return to step 1930.

Secondly, the change request for the type of figure is generated by selecting the item 831 displayed in the figure type designation area 830 with the pointing device. In this case, the flag corresponding to the type of the graphic instructed in step 1960, that is, the flag indicating whether or not to be the search target is switched.

Thirdly, the change request for the distance to the figure is
It occurs by selecting the item 841 or the numerical value change button 843 in the range designation area 840 with a pointing device, or by selecting the line range designation area 842 or the page range designation area 845 and inputting a numerical value. In the present embodiment, the method of selecting the numerical value change button with the pointing device to increase / decrease the numerical value stepwise is adopted, but it is also possible to adopt the method of moving the knob such as the slider with the pointing device to increase / decrease the numerical value. . In this case, in step 1970, it is determined whether the distance designation method is appropriate. If the designation method is not appropriate, an error message is displayed in step 1975 as in the case of the search term. If so, return to step 1930.

Fourthly, the request for changing the document type is generated by selecting the item 811 displayed in the document type designation area 810 with the pointing device.
In this case, in steps 1980 and 1985, as described above, the document database is identified, the items that can be designated as the search condition are checked for the designated type of document, and the user cannot select the item that cannot be designated as the search condition. After doing so, the process returns to step 1930. These processes are repeated until the search execution process request is acquired. In this example,
The type of document is “newspaper / magazine / catalog” 811 and “technical document” 812, the search terms are “mouse” 821 and “hanten” 822, the type of figure is “photograph” 831 and “figure” 832, and the distance between the figure "Within 20 lines" 84
2 is created as a search condition.

Next, in step 1820, the keyword dictionary is referred to, it is determined whether or not there is a search word for which the different notation should be corrected to a predetermined notation, and if there is a corresponding search word. For example, amend the different notation to the prescribed notation. In this example, “mouse” and “hanten” are also predetermined notations, and therefore correction processing is unnecessary.

Next, in step 1825, the upper word list is set to NULL for keyword expansion processing. Then, the process proceeds to step 1830, and it is determined whether or not there is a homonym significant word in the keyword. Specifically, referring to the keyword dictionary, the word table 41 having the search word "mouse" as a synonym
0 (see FIG. 4) is searched to see if the pointer from the word table to the word table corresponding to the homonymous meaning word is NULL. If there is a pointer to the word table corresponding to the homophone meaning word, the process moves to step 1835, and if not, the process moves to step 1850. In this example, the word table 42
Since the pointer to 0 is not NULL, step 183
Go to 5.

Here, the data structure of the word table will be described. The word table 410 is a pointer to a word table 420 corresponding to a homonym meaning word, a pointer to a lower word management table 411 for managing lower words of the word, and a synonym management table 415 for managing synonyms. And a pointer to the word table 440 corresponding to the upper word of the word. In addition, the lower word management table 411
Is composed of a pointer to the word table 418 corresponding to the lower word of the word and a pointer to the next lower word management table 412. The synonym management table 415 includes a pointer to the first character of the character string 483 corresponding to the word in the character string area 480 and a pointer to the next synonym management table 416. In the word table 410, when the synonym management tables 415, 416, and 417 are traced, the synonyms are “mouse” 483, “mouse” 486, and “mouse” 4.
82, and the upper word traces a pointer 419 from the same table to the upper word table, and further the upper word table 4
40, a synonym management table 445 is followed by “animal” 48
1, the lower word management table 411, the word table 418 corresponding to the lower word, and the synonym management table 419 are traced to be "mus musculus" 493. Further, when the word table 420 corresponding to the homonymous meaning words and the synonym management tables 421 and 422 of the table are traced, the synonyms are “mouth” 484 and “mouse” 486, and the words corresponding to the homonym significant words in the table. By tracing the table 430 and the synonym management tables 435 and 436 of the table, it can be seen that the synonyms are “mouse” 486 and “mouse” 485.

Further, it is assumed that the word table has the same number of meanings as words, that is, the number of upper words. In this example, the upper word of the word table 410 is “animal” 481, the upper word of the word table 420 is “body” 492, and the word table 4
The broader term of 30 is “pointing device” 491.

At step 1835, the superordinate word information of "mouse" is acquired. Specifically, in the word table having “mouse” as a synonym, the pointer 419 from the first word table 410 to the word table of the upper word is followed to search the upper word table 440 to search the upper word. Among the synonyms in the table, the pointer to the character string area is traced from the synonym management table 445 at the head, the character string is acquired, the upper word table 640 (see FIG. 6) is generated, and the pointer to the upper word table is obtained. 614 is connected to the keyword table 610. Here, the high-order word table includes a pointer to the next high-order word table, a search / non-search flag 641 and a high-order word 642 of the keyword. In FIG. 6, for simplification, the character string of the upper word is embedded in the upper word table. The word table 420 corresponding to the homonymous meaning words in the word table 410 is traced, and the pointer from the word table 420 to the word table of the superordinate word is traced to the superordinate word table 45.
Search for 0. An upper word table is generated similarly to the word table 440, and the upper word table 64 that has already been generated is generated.
Connect the pointer from 0. This procedure is repeated until there is no word table corresponding to the homonymous meaning word, and a list of upper word tables is generated. When there is no word table corresponding to the homonym significant word, the pointer from the keyword table to the upper word table is set to NULL.

Based on the list of the upper word table thus generated, the upper word designation window 1100 (see FIG. 11) is displayed on the display device in step 1840 to display a list of upper word candidates. The user selects, using the pointing device, a high-order word of a field to be searched from the high-order words displayed. “Animal” 1111 and “pointing device” 1113 are selected, and markers 1121 and 1122 indicating that these words are selected are displayed. In this embodiment, a method of indicating a selected / non-selected state by displaying a character string surrounded by a rectangle is adopted.
The display color and the background color of the character string may be highlighted, or the button corresponding to the character string may be displayed to change the state of the button.

In step 1845, an upper word list is created based on the upper word selected by the user. In particular,
The search / non-search flags 641 and 645 of the upper word table corresponding to “animal” and “pointing device” are set to “S”, and the search / non-search flag 643 of the higher word table corresponding to “body” is set to “NS”. To do. Where "S"
Indicates that it is selected as a search field, while "N
"S" indicates that it is not selected. In this embodiment, a method of creating a list of broader word tables by substituting the value of the search / non-search flag is used, but a method of having only the broader word table corresponding to the broader word selected by the user as a list Can also be taken. This means is similarly repeated for other keywords.

Next, in step 1850, synonym information is acquired. First, in step 2010 (see FIG. 20), the synonym list is set to NULL, and the process proceeds to step 2020.
It is determined whether or not the superordinate word list is NULL. If the upper word list is NULL, the process moves to step 1855. In this example, the broader word list is not NULL, so step 2
Moving to 030, the synonym information matching the upper word in each upper word table is acquired. In step 2040, it is determined whether there are synonyms. If there are synonyms, step 2
Move to 050. Otherwise, return to step 2020. Since the keyword “mouse” has the superordinate terms “animal”, “body”, and “pointing device”, the process moves to step 2050. In the same step, a synonym table 620 (see FIG. 6) corresponding to “mouse” is generated, and a pointer 612 to the synonym table is set to the keyword table 61.
Connect to 0. Here, the synonym table is a pointer to the next synonym table, a search / non-search flag, and a keyword 6.
21 or its synonym 622. Next, "mouse" as an "animal" is synonymous with "mouse" 622, "rat" 623, "rat" 624, and "mouse" 625.
The synonym table corresponding to is generated and connected to the last synonym table. Of the synonyms of "mouse" as "body" and "mouse" as "pointing device", generate a synonym table for the synonyms different from the synonyms that have already been generated, and add them to the list of synonym tables. to add.

In this embodiment, the upper word designation window 110
Although the method of generating only the synonym table corresponding to the synonyms matching the high-order word designated by 0 is adopted, a method of generating the synonym table regardless of the high-order word can also be adopted.

Also, for the keyword "Hanten" 660, the word table 510 of "Hanten" (see FIG. 5) is searched to see if the pointer from the word table to the word table held as a homonymous meaning word is NULL. in this case,
The pointer to the word table 520 is not NULL, so
It can be seen that "hanten" is a homonym meaning word. First, the synonym of the first word table 510 of “Hanten” is “speckle” 58.
2 and the next word table 520 is “reverse” 581. Based on these, a synonym table of the keyword "Hanten" is generated. Also, two word tables 510 and 5
The pointers to the word tables corresponding to the upper 20 words are all NULL, and the pointers to the lower word management table are all NULL.
The pointers from the 60 keyword table to the upper word table and the pointer to the lower word table are both NUL.
It is L. Here, when selecting a synonym to be searched as a keyword, when the user selects a corresponding menu item from the displayed function list, a list of homonyms is displayed. This is a display format similar to that of the upper word designation window 1100, and a method of displaying a homonym meaningful word instead of the upper word may be adopted. The search / non-search flag is set to "S" in the synonym table corresponding to the homonym significant word selected in that window. In this way, "Henten" 661,
A synonym table corresponding to “speckle” 662 and “reverse” 663 is generated, and a list of synonym tables is generated.

After the synonym list is generated for each keyword in this way, the process moves to step 1855 to search for a document containing the keyword or its synonym. Next, the process proceeds to step 1860, and the keyword in the retrieved document is collated with the figure. First, in step 2110 (see FIG. 21), the keyword information in the document is acquired, in step 2115, the searched information table 720 (see FIG. 7) corresponding to the keyword is generated, and the pointer 716 from the document table 710 is connected. The searched information table is composed of a pointer to the next searched information table, a pointer to the corresponding word table, a pointer to the graphic instruction word table, and the like. In step 2120, the graphic directive list is set to NULL. Next, in step 2130, a graphic designator near the keyword is searched, and in step 2140, it is checked whether or not there is a constraint condition regarding the type of graphic among the search conditions. If there is a constraint condition, the process moves to step 2150 to search for keywords near the graphic designator corresponding to the graphic of the specified type among the keywords searched for in step 1855. In this example, since the photograph and the figure are the types of figures to be searched, the figure correspondence table is referred to narrow down the candidates of the figure designator corresponding to these, and then the figure is searched. Unless otherwise specified, the distance between the keyword and the figure directive is within the range of the distance set by default. In the same step,
When there is a pair of the corresponding keyword and figure designating word, the corresponding word table and the figure designating word table are generated according to the contents, and the pointers from the searched information table are connected respectively. Then, the keyword is stored in the corresponding word table, and the graphic indicator is stored in the graphic indicator table. Even if the same figure-directing word appears a plurality of times within a distance from the keyword for a certain keyword, only one figure-directing word table is generated. In this example,
“Document # 40253” 711, “Document # 72884” 7
Let 51 be the document retrieved by the text search,
In “Document # 40253” 711, the keyword near the graphic designator of the photograph or figure is “rat” 731.
And “speckle” 732, the graphic indicator is “photograph 3” 741,
For example, "Fig. 28" 742, and in "Document # 72884", the keywords are "mouse" 761 and "reverse" 76.
2. The figure designator is "Fig. 1.3" 772, "Fig. 2.1" 7
74. On the other hand, step 2160 is the same as step 2150 except for restrictions on the type of figure. In this way, the graphic instruction word list is generated.

Next, in step 2170, the searched graphic list in the graphic designator table is set to NULL, and step 21
Moving to 80, the narrowing down process is performed according to the search condition. In step 2210 (see FIG. 22), the graphic directive list is NU.
It is determined whether it is LL, and if it is not NULL, step 22
Move to 20. If it is NULL, the procedure moves to step 1870 (see FIG. 18). In this example, the graphic directive list is NU
Since it is not LL, the process moves to step 2220. In the same step, it is checked whether or not there is a constraint condition regarding the distance to the figure. If there is a constraint condition, the process proceeds to step 2230, and if there is no constraint condition, the process proceeds to step 2250. In this example, since the distance to the figure is set, the process moves to step 2230. In the same step, the figure corresponding table is identified by referring to the figure correspondence table, the layout information is also referred to determine the positional relationship between the keyword and the figure in the document, and the figure is set as the search condition. It is determined whether it is within the range of the distance. If it is within the range, the process proceeds to step 2240, and the flag within the range of the graphic directive word table corresponding to the graphic directive word is set, and further the graphic flag in the document table connected to the graphic directive word table is set. In this example,
For the in-range flag, "IN" 742 (see FIG. 7) is in the range, and for the figure flag, "T" 71
2 indicates that there is a corresponding figure in the document. On the other hand, step 2250 is the same as step 2230 except for setting the distance to the figure.

Then, the process proceeds to step 1870, and it is checked whether or not there is a document that meets the keyword and the search condition. Specifically, in step 1870, in the document table (see FIG. 7), the total number of documents having a graphic flag, which is “T” in this example, is counted. If the total number is 1 or more, That is, if there is a document that meets the keyword and the search condition, the process moves to step 1880 to display the search result. If there is no corresponding document, step 18
Moving to 85, a message is displayed. In this example, since there is a corresponding document, the process moves to step 1880.

Next, in step 2310 (see FIG. 23), the search result display area 8 in the figure search window 800 is displayed.
The total number of documents is displayed at 50 corresponding positions. Next, proceeding to step 2320, a graphic in the document table for which both the graphic flag and the flag within the range of the graphic designator table are set, that is, the searched graphic is displayed in the same display area, and the user waits for input. In the present embodiment, the method of displaying the retrieved graphic after reducing or enlarging it to a predetermined size is used, but a method of cutting out a part of the graphic to a predetermined size and displaying it can also be adopted.

The user input information is acquired, and step 23 is performed.
At 30, it is determined whether the input information is a display request for a graphic. If the display request is for a figure, step 23
Moving to 40, the document including the figure is displayed (see FIGS. 9 and 10). FIG. 9 is a document including the searched figure 855, FIG.
0 is an example in which a document including the searched graphic 856 is displayed.
FIG. 9 shows a document display window 900 that displays a page including a graphic of a graphic relating to the synonym “rat” 924 of “mouse” and the synonym “speckle” 921 of “hanten”. In the title display area 910 at the top of the window, the document name “Introduction to Pathology” 911 of the document including the graphic and the corresponding page number “page 12” 912 including the graphic are displayed. In the document display area 920, the searched figure 930, among the texts in the displayed document, the synonym “rat” 924 for the keyword “mouse”, the synonym “speckle” 921 for the keyword “hunten”, Graphic instruction word "Photograph 3" 922, 923
Are surrounded by rectangles. In this example, the graphic instruction word “photograph 3” appears in two places, because both are within the range of the distance from the keyword. In the present embodiment, the method of displaying the keyword or its synonym and the graphic designating word by enclosing them in a rectangle is adopted, but a method of displaying by inverting the display color and the background color of the character can also be adopted. . There is also a method of starting an application that created the document and displaying the document according to a user's instruction.

On the other hand, if it is determined in step 2330 that the user's input information is not a display request for a graphic, step 18
The process returns to 05 and waits for user input.

If the input information of the user is acquired and the input information is a request for ending the session, the process proceeds to step 1890, and if not, the process proceeds to step 1810. here,
If the input information is not a session end request, the search word, search condition, and search result information up to this point are saved in another area of the memory, and then the user waits for input. The subsequent method of editing the search condition, the method of searching, and the method of displaying the search result are the same as before. Step 180
In step 5, if the input information is a session end request, the process moves to step 1890, and the user is asked whether to save the search result information during the session. When the input information corresponding to the storage processing request is acquired, in step 1895,
The information in the keyword table, document table, and searched information table is saved in a file on the computer.

As described above, the search condition can be edited, the graphic can be searched, and the search result can be displayed.

Next, as a second embodiment of the search method of the present invention, a method of expanding a lower word will be described. In this example, it is assumed that the search word “mouse” 821 is displayed in the search word input area 820 of the graphic search window 800 (see FIG. 8). When the user selects the search word “mouse” 821 with the pointing device and selects the corresponding menu item from the displayed function list, the subordinate word expansion window 1200 (see FIG. 12) is displayed on the display device. The lower word expansion window is displayed in the upper word display area 121.
0, lower word display area 1220 and lower word input area 123
It consists of 0. If the search word has an upper word, the upper word display area 1210 displays a list of the upper words. Specifically, the word table 410 (see FIG. 4) having the search word “mouse” as a synonym is searched by referring to the keyword dictionary,
It is checked whether the pointer to the word table corresponding to the high-order word is NULL. In this example, the pointer to the word table corresponding to the high-order word is not NULL, so the pointer 419
To obtain information on the higher-ranked word "animal". Next, the pointer to the word table 420 of the homonymous meaning word and the pointer to the word table 450 corresponding to the higher word thereof are traced to obtain the information of the higher word “body”. Similarly, the information of another higher-ranking word "pointing device" is also acquired. Based on the information of the higher-ranked words thus obtained, a list of higher-ranked words of the search word is displayed in the higher-ranked word display area. The user can select a broader word displayed in the same area with the pointing device. The selected high-order word is displayed surrounded by a rectangle 1215.

The lower word display area 1220 displays a list of lower words corresponding to the search word having the meaning of the upper word selected in the upper word display area. In this example, the user checks the pointing "animal" 1211 from the word table 410 corresponding to the search word "mouse", that is, the pointer from the word table corresponding to "mouse" as an animal to the lower word management table is NULL. In this example, since the pointer to the subordinate word management table is not NULL, the pointer is traced, and further the pointer from the subordinate word management table 411 to the word table 418 corresponding to the subordinate word is traced to acquire the information of the subordinate word "mus musculus". To do. Similarly, the subordinate word information of "mouse" as an animal is acquired. The lower word thus obtained is displayed in the lower word display area.
Also, the selected subordinate word, "mus musculus" in this example 1
221 is displayed surrounded by a rectangle 1225.

The user can add a lower word by inputting a character string in the lower word input area 1230. In this example, enter "Hamster" 1231,
When the corresponding menu item is selected from the displayed function list, the character string is additionally displayed in the lower word display area. When the user wants to reuse the subordinate word input here, the user selects the corresponding menu item from the displayed function list to register the subordinate word in the keyword dictionary as a subordinate word of "mouse" as an animal. You can also

After the user selects a subordinate word displayed in the subordinate word display area with the pointing device, and selects the corresponding menu item from the displayed function list, the subordinate word of the search word is also searched. Can be added as a text search. In this embodiment, when the word displayed in the upper word display area and the lower word display area is selected, the method of displaying the character string by enclosing it in a rectangle is adopted. It is also possible to adopt a method of displaying the color in reverse, displaying a button around the character string, and displaying the display color and the background color of the button in reverse.

When the keyword table of the search word is already generated, by setting the mode so that the lower word is expanded at the time of search, the lower word information selected in the lower word expansion window can be added. it can. Specifically, the subordinate word information selected in the subordinate word expansion window, in this example, the subordinate word table 630 (see FIG. 6) corresponding to the subordinate word "mus musculus" of "mouse" as an animal is generated, and the keyword " A pointer 613 from the keyword table 610 corresponding to "mouse" to the lower word table is connected. Next, a subordinate word table corresponding to "guinea pig" is generated and connected to the subordinate word table of "mus musculus". Further, the search / non-search flag of the lower word varies depending on whether it is selected in the lower word expansion window.
In this example, "mus musculus" 631 has been selected in the same window, so the same flag means "S" for search.
On the other hand, since "guinea pig" 632 is not selected, it can be seen that it is "NS" meaning non-search.

In the present embodiment, the method of displaying all the upper words of the search word in the upper word display area is adopted. However, when the keyword table of the search word is generated, the search / non-search of each upper word table is performed. Refer to the value of the flag, and that value is
It is also possible to adopt a method in which only the high-ranking word that is “S” meaning search for that word can be selected or only the high-ranking word is displayed. In this example, "animal" 642
(See FIG. 6) and the search / non-search flags 641 and 645 of the “pointing device” 646 are “S”, and the flag 643 of the “body” 644 is “NS”, so that the “animal” 1221 and the “pointing” “Device” 1223 can be selected. And the "body" 122
Either 2 cannot be selected or it is not displayed in the upper word display area. In the present embodiment, a method of setting a mode for expanding a lower word at the time of searching is adopted, but a method of expanding a lower word can also be adopted as a preprocessing of text search. As described above, the expansion of the lower word can be performed.

Next, as a third embodiment of the search method of the present invention, a method of designating the position of a figure on a page will be described. In this example, the figure search window 800 (see FIG.
(See) is displayed on the display device. When the user selects the corresponding menu item displayed in the function list display area 860, in this case, the search menu 861 with the pointing device, the position designation window 1300 (see FIG. 13) is displayed on the display device. The position designation window is made up of an item “no designation” 1310 indicating no position designation and an item specifically indicating the position on the page of the graphic. In this example, "upper left" 1
321, "upper right" 1322, "left center" 1323, "right center" 1324, "lower left" 1325, "lower right" 1326
6 items. Each item has a button indicating the selected / non-selected state of the item. In FIG. 13, “upper left” 132
1, "upper right" 1322, "left center" 1323, and "right center" 1324 are in the selected state, which indicates that the figure at the top or center of the page is designated. When "unspecified" is in the selected state, when the user selects any of the items indicating the position within the page, "unspecified" is in the unselected state.

In FIG. 13, the page is divided into 6 areas, that is, the page is divided into three in the vertical direction and two in the horizontal direction to specify the position, but the number of page divisions is interactively set. However, it is also possible to display the divided areas in a grid and specify the area of the grid by the user with the pointing device, or the method in which the user draws a rectangle on the screen with the pointing device. it can. For example,
In FIG. 14, when the user selects the item “designated” 1412 with the pointing device, the item is in a selected state, and the vertical division number input area 1413 and the horizontal division number input area 1414 for designating the number of divisions of the page. Can be input. After the user inputs an appropriate number in these input areas, by selecting the corresponding menu item from the displayed function list, the grid display area 14 is displayed.
20 is displayed in a grid pattern. The user can specify the position of the figure on the page by selecting the lattice with the pointing device. Further, for example, in FIG. 15, when the user selects the size of the document to be searched from the list of paper sizes displayed in the paper size designation area 1510,
The mount 1530 and the cursor 15 in the position designation area 1520
40 is displayed. In this case, in order to make it clear that the position is being designated, the cursor shape is changed and displayed. Here, the user can specify the position of the figure within the page by drawing a rectangle on the mount with the pointing device.

After designating the position within the page in this way, the layout information is referred to search for a graphic that also satisfies other search conditions. This designation method is effective when searching for a figure that remembers the rough position on the page. Figure 10
An example of displaying a document including a figure retrieved by designating an in-page position is shown in FIG. The figures 1030 and 1040 are keywords, and in this example, the distances from the “mouse” 1022 and the “reverse” 1023 are within the search condition range. In this case, the figure 1030 is within the specified range of the in-page position. Since the figure 1040 is outside the specified range, the figure 10
Only 30 is given to the user as a search result. In the present embodiment, a method of narrowing down the search only to the graphics within the range of the area designated by the position designation window is used, but the graphics protruding outside the area of the range are also targeted for the search. , The user sets an appropriate threshold value, adds the width of that value to each area, and expands the area.
It is also possible to adopt a method of searching for a figure in the expanded area. As described above, the position of the figure within the page can be specified.

Next, as a fourth embodiment of the retrieval method of the present invention, the correspondence between keywords and figures will be described. In this example, the figure search window 800 (see FIG.
It is assumed that the searched figure is displayed in the search result display area 850 (see the reference). When the user selects a figure to be stored by associating with a keyword or the like from those figures with the pointing device and selects the corresponding menu item from the displayed function list, the figure registration window 1600 (see FIG. 16) is displayed. Display on the device.
The figure registration window includes a search word display area 1610, a corresponding keyword display area 1620, a corresponding document name display area 16
30 and a label input area 1640. In the search word display area, the same search word as the search word displayed in the search word input area 820 of the figure search window is displayed, and in the corresponding keyword display area, the keywords corresponding to the searched figure are displayed in a list. The document name of the document including the graphic is displayed in the document name display area. In the label input area, a label name indicating the correspondence between the keyword and the graphic is input. In addition to entering the label name from the keyboard, specify the character string or part of the character string displayed in each area of the search word display area, the relevant keyword display area, and the document name display area with the pointing device to input the label. You can also copy to the area. In this example, the search word "mouse" 1
611 and the corresponding keyword “speckle” 1622 are specified,
Copying to the label input area.

Then, by selecting the corresponding menu item from the displayed function list, the label thus input is registered in the graphic correspondence table together with the search word, the keyword and the like. As a result, it is possible to search for a desired figure by designating the label of the figure once searched. Further, when displaying the search result, it is also possible to preferentially display the figure associated with the keyword or the like. As described above, the keyword and the like can be associated with the figure.

Next, as a fifth embodiment of the search method of the present invention, a search result narrowing method will be described. In this example, it is assumed that the figure search is performed multiple times. When the user selects the corresponding menu item with the pointing device from the displayed function list, the narrowing-down specification window 1700 (see FIG. 17) is displayed on the display device. The narrowing-down specification window has a search word set display area 1710. In the same area, a list of the search terms specified in each search and edit process and the number of found graphics is displayed in the order in which the graphics search was executed. In this example, "synonyms", "patents", and "search" are the first search word sets, and "mouse" and "hanten" are the subsequent search word sets. In the present embodiment, a method of displaying the search word set and the number of search results in the search word set display area is adopted, but a method of displaying the search condition such as the document type can also be adopted.

After the user selects a plurality of search word sets with the pointing device and selects the corresponding menu item from the displayed function list, the set of search results to be narrowed down can be specified. The search results can be narrowed down as described above.

Next, a sixth embodiment of the present invention will be described.
Currently, an image retrieval system that retrieves a desired image based on a drawing drawn by a user has been proposed (Kyoji Hirata et al. "Image retrieval using rough images", IEICE Technical Report Vol.92, No.51, DE92-2,
1992.5). In this image retrieval system, features such as shape, position, color and structure in an image are automatically extracted when a database image is stored and a schematic image is created. Next, the created schematic image and the drawing drawn by the user (referred to as “inquiry image”) are matched, and a desired image is searched using the result. Since a rough image is created, it is possible to search even if the "inquiry image" is ambiguous.

However, in this image retrieval system, since the "inquiry image" drawn by the user is rough and ambiguous, a large number of images similar to the "inquiry image" are output even if they are not the desired images. In addition, since a schematic image must be stored for each original image, the storage amount increases.

Therefore, in the present embodiment, the above problem is solved by paying attention to the feature that a general document includes a figure and a text. First, the figure is searched using the text. That is, as shown in the first embodiment, the desired figure is searched using the search word and the search condition. At this time, the figure retrieved is not always limited to one, so the figure is narrowed down based on the figure drawn by the user (called a rough sketch in this embodiment). For the problem of increasing the storage amount, it is only necessary to extract the features of the searched graphic using the search word and the search condition, instead of extracting the characteristics of all the accumulated graphics.

FIG. 24 shows a display state in this embodiment.
In this case, five figures were searched using the search word and the search condition. A figure is drawn in the drawing area 2410, and five figures are narrowed down based on this figure. The result is the searched figure 855.
Is output. Here, although the search is performed based on the search word and the search condition and then narrowed down based on the rough sketch, the search may be performed using both the search using the search word and the search condition and the search using the rough sketch. Therefore, the rough sketch may be searched first, and then the search may be narrowed down by the search word and the search condition. Also, 2
It is also possible to perform one search separately and take the AND of the results.

Further, if the user can directly select one from the retrieved figures or input a rough sketch, rather than automatically performing the narrowing down using the rough sketch, the retrieval efficiency will be improved. Go up. Further, the number may be narrowed down when the number of retrieved figures is a certain number or more. As a method of determining the fixed number, it may be determined whether or not all the retrieved figures are included in one display screen.

[0064]

According to the present invention, it is possible to efficiently search for a figure, table, photograph or the like included (or linked) in an electronic document from a word input by a user. Specifically, by referring to the keyword dictionary, a synonym expansion of the input keyword is performed, a text search is executed, and then the layout information of the electronic document is referenced to search for a graphic in the vicinity thereof. Further, it is possible to search for a graphic by using the positional relationship between the graphic and the keyword, the position of the graphic within the page, and the like as search conditions. By expanding the subordinate words of the keyword according to the meaning of the keyword and searching for all or a part of the subordinate words, the number of keywords input by the user is reduced, and even a user who is not familiar with the keyword classification system can easily perform it. Since a keyword can be specified in, it is possible to search even from ambiguous keywords and reduce the burden on the user.

[Brief description of drawings]

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

FIG. 2 is a system configuration diagram of the present invention.

FIG. 3 is a diagram showing a flow of processing at the time of graphic search.

FIG. 4 is a diagram showing an example of a data structure of the present invention.

FIG. 5 is a diagram showing an example of a data structure of the present invention.

FIG. 6 is a diagram showing an example of an internal data structure of the present invention.

FIG. 7 is a diagram showing an example of an internal data structure of the present invention.

FIG. 8 is a diagram showing a screen on which a search result and a search condition are edited and a search result is displayed.

FIG. 9 is a diagram showing a screen displaying a document including a searched figure.

FIG. 10 is a diagram showing a screen displaying a document including a searched figure.

FIG. 11 is a diagram showing a screen for selecting an upper word of a search word.

FIG. 12 is a diagram showing a screen for selecting a lower word of a search word.

FIG. 13 is a diagram showing a screen for designating the position of a figure on a page.

FIG. 14 is a diagram showing a screen for designating the position of a figure on a page.

FIG. 15 is a diagram showing a screen for designating the position of a figure on a page.

FIG. 16 is a diagram showing a screen for inputting a graphic label name.

FIG. 17 is a diagram showing a screen for designating a search word set.

FIG. 18 is a diagram showing a processing flow for creating a search condition and searching for a figure.

FIG. 19 is a diagram showing a processing flow for creating a search condition.

FIG. 20 is a diagram showing a processing flow for acquiring synonym information.

FIG. 21 is a diagram showing a processing flow of collating a keyword with a figure.

FIG. 22 is a diagram showing a processing flow of narrowing down.

FIG. 23 is a diagram showing a processing flow of displaying a search result.

FIG. 24 is a diagram showing a display state of a sixth embodiment.

[Explanation of symbols]

101 ... Keyboard, 102 ... Pointing device, 103 ... Display device, 110 ... Interactive display unit, 120 ...
Dialog control unit 121 ... Input analysis unit 122 ... Window / menu control unit 130 ... Search condition display control unit 13
1 ... Search condition creation unit, 132 ... Search result display unit, 150
... graphic search unit, 151 ... keyword expansion unit, 152 ... graphic matching unit, 155 ... document search unit, 160 ... document management unit,
170 ... Keyword dictionary, 175 ... Graphic correspondence table, 180
... Allocation information, 185 ... Multimedia document database, 201 ... Control device, 205 ... Main storage device, 206 ...
Magnetic disk group, 410 ... Word table, 411 ... Subordinate word management table, 415 ... Synonym management table, 480 ...
Character string area, 610 ... Keyword table, 620 ... Synonym table, 630 ... Lower word table, 640 ... Higher word table, 710 ... Document table, 720 ... Retrieving information table, 730 ... Applicable word table, 740 ... Graphic instruction word Table, 800 ... Graphic search window, 810 ...
Document type designation area, 820 ... Search word input area, 830 ...
Graphic type designation area, 840 ... Range designation area, 842 ... Line range designation area, 845 ... Page range designation area, 850 ...
Search result display area, 855 ... Searched figure, 860 ... Function list display area, 900 ... Document display window, 910
... Title display area, 911 ... Document name, 912 ... Corresponding page number, 920 ... Document display area, 921 ... Corresponding keyword, 922 ... Graphic instruction word, 930 ... Searched figure, 11
00 ... Higher word designation window, 1111 ... Higher word, 12
00 ... lower word expansion window, 1210 ... upper word display area, 1211 ... upper word, 1220 ... lower word display area, 1
221 ... Lower word, 1230 ... Lower word input area, 1300
... Position designation window, 1413 ... Vertical division number input area, 1414 ... Horizontal division number input area, 1420 ... Lattice display area, 1510 ... Paper size designation area, 1520 ...
Position designation area, 1530 ... Mount, 1531 ... Designation range,
1540 ... Cursor, 1600 ... Graphic registration window,
1610 ... Search word display area, 1620 ... Applicable keyword display area, 1630 ... Applicable document name display area, 1640 ...
Label input area, 1700 ... Narrow down designation window,
1700 ... Search word set display area, 2400 ... Sketch input window, 2410 ... Drawing area, 2411 ... Rough sketch.

Claims (18)

[Claims] 1. An electronic document composed of text and graphics is stored in advance, and at least one search word for searching a predetermined electronic document from the stored electronic document is input, and the input search word is input. And interactively edit the search condition that specifies the positional relationship between the desired figure and the electronic document that includes the entered search term, and retrieves the figure corresponding to the edited search condition. A fuzzy figure retrieval method characterized by identifying the figure from the figure of the document and writing the identified figure. 2. The ambiguous figure search method according to claim 1, wherein electronic documents composed of texts and figures are stored in advance for each document type, and at least one document type to be searched is selected from the stored electronic documents. An ambiguous figure retrieval method comprising selecting and identifying a desired figure from the selected document type. 3. The ambiguous figure search method according to claim 1, wherein at least one type of figure to be searched is designated to reduce the total number of figures to be searched and electronic documents. retrieval method. 4. The fuzzy graphic search method according to claim 2, wherein the search condition quantitatively designates a neighborhood condition between the search word for searching in the text and a desired graphic by the number of pages and the number of lines. A fuzzy figure retrieval method characterized by setting by doing. 5. The ambiguous figure searching method according to claim 2, wherein the total number of figures to be searched is reduced by designating a position of a desired figure in a page. 6. The ambiguous figure search method according to claim 1, wherein the presence or absence of a homonym of the input search word is determined, and if the input search word has a homonym, the input A fuzzy figure search method characterized in that a search word is displayed as an abstract word, and at least one of the displayed upper words is selected to narrow down the field to be searched. 7. The fuzzy graphic search method according to claim 6, wherein an appropriate synonym is expanded in accordance with a broader term of the input search word, and the expanded synonym is stored in advance as a keyword. A fuzzy figure search method characterized by searching text in a document. 8. The fuzzy graphic search method according to claim 6, wherein an appropriate lower word is expanded according to an upper word of the input search word, and the search word and the expanded lower word are used as keywords in advance. A fuzzy figure retrieval method characterized by retrieving text in a stored electronic document. 9. The fuzzy graphic search method according to claim 1, wherein the distance between the search word and a graphic designating word designating a graphic such as a figure number or a table number is quantitatively set, and the search word or graphic The pre-stored electronic document including the directives is searched to generate a list of word sets composed of search words and graphic directives within the set distance, and the graphic instructions of the generated list are generated. A fuzzy figure retrieval method characterized in that a figure close to a keyword is searched by identifying a figure corresponding to a word. 10. The fuzzy graphic search method according to claim 9, wherein the list of generated word sets is narrowed down to a set of words having a graphic designator corresponding to a type of graphic. Method. 11. The fuzzy graphic search method according to claim 9, wherein a distance between the search word and the graphic is set in advance, and the search word and the graphic that meet the set distance are identified. Method. 12. The fuzzy graphic search method according to claim 9, wherein the graphic is identified by referring to pre-stored allocation information and searching for a document having a graphic that satisfies the positional condition of the graphic within the page. An ambiguous figure retrieval method characterized by: 13. The fuzzy figure retrieval method according to claim 1, wherein the retrieved figure is displayed in an enlarged or reduced form or a part of the figure. 14. The fuzzy graphic retrieval method according to claim 13, further comprising means for selecting at least one graphic from the displayed graphic and displaying a corresponding page of a document including the selected graphic. A characteristic fuzzy figure search method. 15. The ambiguous figure retrieval method according to claim 1, wherein at least one figure is selected from the displayed figures, and at least one search term is selected from the search terms that match the selected figure. Then, the fuzzy figure retrieval method is characterized by storing the correspondence between the selected figure and the selected search term. 16. A plurality of electronic documents composed of texts and figures are stored in advance, and at least one search word for searching a predetermined electronic document from the plurality of stored electronic documents is input, and the input is input. Interactively edit the search condition that specifies the positional relationship between the search word and the desired figure, search for an electronic document that includes the input search word, and edit the edited electronic document from the searched electronic documents. The first figure candidate is searched based on the search condition, the outline of the desired figure is drawn, and the second figure candidate is searched from the accumulated electronic documents based on the drawn outline of the figure. Then, the ambiguous figure retrieval method is characterized in that a desired figure is identified based on the first and second figure candidates, and the identified figure is displayed. 17. Storage means for storing an electronic document composed of text and graphics in advance; input means for inputting at least one search word for searching a desired graphic from the stored electronic documents; Edit means for interactively editing a search condition that specifies the positional relationship between the search word input by the input means and a desired figure; and text in an electronic document containing the search word input by the input means. A searching means for searching; an identifying means for identifying a figure corresponding to the search condition edited by the editing means from the figures of the electronic document searched by the searching means; and an identifying means for identifying the figure. An ambiguous figure retrieval device having a display unit for displaying the figure. 18. A storage unit for storing a plurality of electronic documents composed of texts and figures in advance, and at least one search word for searching a predetermined electronic document from the plurality of electronic documents stored in the storage unit. Search word input means, edit means for interactively editing search conditions that specify the positional relationship between the search word input by the search word input means and a desired figure, and input by the search word input means An electronic document search means for searching an electronic document including the searched word, and a desired figure from among the electronic documents searched by the electronic document search means based on the search condition edited by the editing means. A first graphic search means for searching for a desired graphic, a graphic input means for inputting a general shape of a graphic for searching for a desired graphic, and the accumulation based on the general shape of the graphic input by the graphic input means. Electricity accumulated in the means Second figure searching means for searching a figure from a document, identifying means for identifying a desired figure based on the search results of the first and second figure searching means, and a figure identified by the identifying means An ambiguous figure retrieval device having means for displaying.