JP2001014326A - Device and method for retrieving similar document by structure specification - Google Patents

Abstract

PROBLEM TO BE SOLVED: To add the specification of an object structure to be retrieved to retrieval conditions and to improve the retrieval precision when a document which is similar to a seed document (document specified as a retrieval condition) is retrieved. SOLUTION: A retrieval condition expression analyzing program 130 receives the specification of a seed document and the input of an object structure to be retrieved as retrieval conditions. A featured character string extracting program 150 extracts a featured character string from the text of the specified seed document. A retrieval object structure ID acquiring program 151 converts the specified structure into its ID. A similarity calculating program 152 performs retrieval from an appearance frequency file 181 to acquire the appearance frequency of a document whose structure ID matches the featured character string and calculates the similarity of the similar document based upon the seed document. A retrieval result output program 132 displays the identifier and similarity of the similar document as the retrieval result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for searching for a document similar to a document (seed document) specified as a search condition, and more particularly to an apparatus and a method for searching a structured document structure. About.

[0002]

2. Description of the Related Art In recent years, with the spread of personal computers and the Internet, the number of digitized documents stored in a database has been increasing. Search with high accuracy,
There is an increasing demand for fast and efficient searches.

[0003] Various search techniques have been proposed to meet such demands. For example, according to Japanese Patent Laid-Open No. Hei 10-240752, a condition relating to a logical structure is added to a search condition for a document (hereinafter, referred to as a structured document) in which individual logical structural elements constituting the document can be identified. By performing the search, a highly accurate search can be performed.

Japanese Patent Application Laid-Open No. 11-143902 discloses that
Disclosed is a similar document search technique in which a user specifies a document or a sentence having desired contents (hereinafter, referred to as a seed document) and searches for a document similar to the document. According to this technique, a target document can be easily searched for only by showing a sample document, and the user does not need to think or input a complicated search condition expression, and can perform an efficient search.

[0005]

The above-mentioned JP-A-11-1
According to the technique disclosed in Japanese Patent No. 43902, the user can easily and efficiently perform a search, but has a problem of search accuracy as exemplified below.

FIG. 3 is a diagram showing a processing procedure of a conventional similar document search system. The search condition acquisition program displays a guidance screen for inputting search conditions on a display device. For example, list information such as document numbers and headings of a plurality of candidate documents including a seed document is displayed. When a seed document is specified as a search condition, a feature n-gram extraction program is started, the entire text of the seed document is extracted from the document file, and a characteristic character string is extracted from the text. Next, the similarity calculation program is started, the document number and the number of appearances of the characteristic character string are registered corresponding to the characteristic character string, and the appearance frequency file is referred to, and the same characteristic is determined based on the characteristic character string of the seed document. The similarity of the related document using the character string to the seed document is calculated, and list information such as the document number, similarity, and heading of the candidate document is displayed as a search result.

According to the search results shown in FIG. 3, although the document which is very similar to the seed document is the document 4, the similarity of the less related document 1 is higher because the appearance frequency of the characteristic character string is higher. There is a problem that it becomes expensive and is displayed preferentially.

An object of the present invention is to provide an apparatus and a method for searching for a similar document which avoids inputting complicated search conditions but has a high search accuracy.

[0009]

SUMMARY OF THE INVENTION The present invention relates to a method for retrieving a structured document similar to a seed document using a computer, wherein at least the seed document and the structured document belonging to the structured document are used as search conditions for similarity calculation. The method is characterized by a method of searching for similar documents by designating a structure, comprising a step of receiving designation of one structure, and a step of giving priority to a target document having a higher similarity after calculating the similarity.

According to another aspect of the present invention, there is provided a method for retrieving a document similar to a structured seed document, comprising the steps of receiving a designation of a seed document and at least one structure belonging to the seed document; And displaying a target document having a higher similarity with priority.

Further, the present invention is characterized by a retrieval device having the above functions.

Here, designating the structure means designating the names of the logical structural elements constituting the document.

[0013]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a configuration diagram of a similar document search system according to the first embodiment. The computer hardware that implements this system includes a display device 100, an input device 101, a central processing unit (CPU) 102, an external storage device 103, a floppy disk drive (FDD) 104, a main memory 1
06 and a bus 107 connecting these devices.

The external storage device 103 stores a text 180,
An appearance frequency file 181 and a structure index 182 are stored. The text 180 stores a structured document file or a set of unstructured document files.
Here, the structured document is a document having a logical structure conforming to a standard format such as SGML, XML, or a plurality of flat texts extracted for each logical structure. The document stored in the floppy disk 105 via the FDD 104 is registered in the text 180 via the main memory 106.

The system control program 110 stored in the main memory 106 includes an operating system, a program for providing a graphical user interface, and the like. The document registration control program 111 controls execution of a document registration program. The registration program includes an appearance frequency file creation program 1 including a text registration program 120 and an appearance frequency counting program 140.
21 and a structure index creation program 122.
The text registration program 120 is a program for registering a document on the floppy disk 105 in the text 180.

The search control program 112 is a program for controlling execution of a program related to searching for similar documents. The search programs include a search condition expression analysis program 130, a similar document search program 131, and a search result output program 132. The similar document search program 131 includes a characteristic character string extraction program 150, a search target structure ID acquisition program 151, and a similarity calculation program 152. The functions of these search programs will be described in the following description of the search procedure. The search control program 112 and the search program can be stored in a storage medium, read into the main memory 106 via a drive device, and executed by the CPU 102.

As a shared library 160 stored in the main memory 106, there is a structured document analysis program 170. The work area 161 is an area used as a temporary storage area or a work area for data read from the text 180, the appearance frequency file 181, and the structure index 182.

The appearance frequency file 181 stores a document number, an ID of a logical structure including the character string, and the number of appearances in the logical structure corresponding to a character string or a word as shown in a part of FIG. I do. The target document registered in the appearance frequency file 181 is a structured document or an unstructured document. The appearance frequency file creation program 121
The document in the text 180 is read one by one, a characteristic character string is extracted from the text, and the appearance frequency counting program 1
40, the number of appearances of the characteristic character string for each logical structure is counted, and an appearance frequency file 181 is created and registered in the external storage device 103. For a document that is not structured, when a logical structure category is specified, the characteristic character string is associated with the logical structure ID according to the specification. For example, JP-A-11-143902 discloses a processing procedure of an appearance frequency file creation program.

The structure index 182 stores the correspondence between a logical structure and its ID as shown in a part of FIG. The structure index creation program 122 calls the structured document analysis program 170, analyzes the logical structure of the document text read from the text 180, assigns an ID to each logical structure, and registers the logical structure in the external storage device 103. For example, Japanese Patent Laying-Open No. 10-240752 discloses a processing procedure of a structure index creation program.

FIG. 2 is a diagram showing a processing procedure of the first embodiment. In the first embodiment, it is assumed that a seed document is an unstructured document and a search target document is a document (structured document or unstructured document) registered in the appearance frequency file 181. The search condition expression analysis program 130 displays a guidance screen on the display device 100 and accepts an input of a search condition expression.

Here, the search condition expression is a seed document and at least one search target structure. The seed document can select one of a plurality of document candidates in which a heading or the like is already displayed, can directly input the seed document from the input device 101, or can input the FDD 104 or the CD-ROM device ( It is also possible to input via a network (not shown), a network (not shown), or the like.

Further, as shown in FIG.
A search condition expression may be input to 0 through a screen interface including a seed document input area 1200, a search target structure input area 1201, and a search execution button 1202. In the seed document input area 1200, a seed document can be directly input from the input device 101.
Alternatively, the text on the search result display screen (not shown) can be copied to the seed document input area 1200. Alternatively, the seed document may be the indicated portion of such text.

At least one search target structure can be selected from a displayed drop-down menu. In a search condition when a plurality of search target structures are specified, it is possible to assign a weight to each structure. Here, the weight may be input via a weight input area (not shown) or may be defined in a system definition file (not shown).

When the seed document and the search target structure are input via the input device 101, the search condition analysis program 13
0 acquires the text of the seed document from the specified search condition. Instead of specifying a structure to be searched, a structure not to be searched (a structure excluded from the search) may be specified. In that case, the search condition expression analysis program 130 sets the remaining structure as the search target structure. Instead of inputting the search target structure as one of the search condition expressions via the input device 101, a search target structure set in advance in a system definition file (not shown) may be used.

Search target structure ID acquisition program 151
Acquires the identifier corresponding to the specified structure with reference to the structure index 182 as the search target structure ID. In addition, the characteristic character string extraction program 150
, The full text of the specified seed document is extracted, a characteristic character string is extracted, and the number of appearances of the extracted characteristic character string is counted. As a method of extracting a characteristic character string, for example,
The method described in 1-1143902 can be used. In the example of FIG. 2, the extracted characteristic character strings having the higher priority are used. Alternatively, words may be extracted from the document text, and words may be extracted while referring to a word dictionary (not shown) to check for a match with the registered words.

Next, the similarity calculation program 152 obtains the document number of the document whose structure ID matches the search target structure ID with each characteristic character string or word extracted with reference to the appearance frequency file 181, and its appearance frequency. Next, the appearance frequency file 18
1 and obtains the frequency of occurrence of a character string or word other than the characteristic character string or word extracted for the search target structure ID of each document acquired with reference to No. 1 and calculates the degree of similarity with the seed document for each document I do. As a similarity calculation method, for example, Equation 1 described in JP-A-11-143902 can be used. Alternatively, a feature vector having a normalized appearance weight of each feature character string (word) of the seed document as an element and a feature vector of each acquired document are obtained. The degree may be calculated.

Finally, the search result output program 132
The acquired documents are rearranged in descending order of similarity, the display priority is determined in descending order of similarity, and the document numbers and their similarities are displayed on the display device 100 in descending order of priority. Bibliographic items such as the headline and outline of each document may be acquired with reference to a file (not shown in the figure) and displayed together. The document number, similarity, headline, etc. of the seed document can also be displayed for comparison with a similar document.

When a plurality of search target structures are designated, when calculating the similarity of each document, a cumulative value which accumulates the similarity of each search target structure over the entirety is obtained, and the document is determined by this similarity. May be sorted in descending order of the cumulative value of.
Here, the cumulative value is a value obtained by calculating the sum of squares of the similarity for each search target structure and the square root of the sum of squares. Alternatively, the highest similarity among the similarities of a plurality of search target structures may be adopted for each document, and the documents may be sorted in descending order of the adopted similarity. For example, when the same type of logical structure repeatedly appears in a document as in [claim n] in a patent specification, a similarity is calculated for each logical structure,
If the highest similarity is adopted and compared with the similarity of the seed document, it is possible to compare the logical structures having the highest similarity in content regardless of the order of the logical structure of the same type. Further, both a mode for obtaining the cumulative value of the similarity and a mode for employing the highest similarity may be provided so that any one of the search conditions can be selected, or a system definition file (FIG. (Not shown in the figure) may be set.

FIG. 4 is a diagram showing a configuration of the similar document search program 131a according to the second embodiment. In the second embodiment, a seed document structure analysis program 400 is added to the characteristic character string extraction program 150a. The seed document structure analysis program 400 is configured to call the structured document analysis program 170 stored in the shared library 160. In addition, a correspondence structure determination program 401 is added to the similarity calculation program 152a.

FIG. 5 is a diagram showing a processing procedure according to the second embodiment. In the second embodiment, the seed document is a structured document, and the search target document is a document (structured document or unstructured document) registered in the appearance frequency file 181. The search condition of the second embodiment is a seed document and at least one structure belonging to the seed document. When the seed document and the structure are input via the input device 101, the search condition expression analysis program 130 acquires the text of the seed document from the specified search condition. It is assumed that the logical structure of the seed document specified by the search condition matches the logical structure of the search target document. Instead of specifying a structure to be searched, a structure not to be searched (a structure excluded from the search) may be specified.
In this case, the search condition expression analysis program 130 sets the remaining structure belonging to the seed document as the search target structure. Similar to the first embodiment, the search target structure may be set in the system definition file in advance.

Next, the seed document structure analysis program 400
The text of the seed document is extracted from the text 180, and the structure of the seed document is analyzed to extract only the body text relating to the specified structure. It is an error if the text of the seed document does not contain the specified structure. As a method of analyzing the structure of a document, for example, a processing procedure of a document structure analysis program is described in Japanese Patent Laid-Open No. Hei 10-240752. Next, the search target structure ID acquisition program 151
Acquires the search target structure ID corresponding to the specified structure with reference to the structure index 182. The characteristic character string extraction program 150a extracts characteristic character strings (words) of the extracted text and counts the number of appearances of the extracted characteristic character strings. For the seed document, the characteristic character string of each structure has already been extracted, the number of appearances for each structure has been counted, and if it has been registered as in the appearance frequency file 181, the file is specified by referring to that file. It is only necessary to extract the structure, the characteristic character string, and the number of appearances. In this case, the processing of the seed document structure analysis program 400 and the characteristic character string extraction program 150a can be skipped.

Next, similar to the first embodiment, the similarity calculation program 152a extracts each characteristic character string (word) extracted with reference to the appearance frequency file 181 and the search target structure ID.
Is obtained, and the similarity with the seed document is calculated for each document. At this time, the corresponding structure determination program 401 determines the structure of the seed document and the appearance frequency file 18
1, the characteristic character strings are divided into groups for each search target structure, and the similarity for each search target structure is calculated. When a plurality of search target structures are specified, the method of calculating the final similarity of each document is the same as in the first embodiment. Finally, the search result output program 132 sorts the acquired documents in descending order of the calculated similarity, and displays the document numbers, similarities, headings, and the like on the display device 100.

In the above description of the second embodiment, it is assumed that the structure of the specified seed document matches the search target structure, but both may have different logical structures. That is, when the structure of the seed document is specified and another search target structure is specified, the seed document structure analysis program 400 and the characteristic character string extraction program 150a pay attention to the structure of the specified seed document. The character string is extracted, and the similarity calculation program 152a searches the search target document by paying attention to the ID of the specified search target structure. The correspondence structure determination program 401 associates the structure of the specified seed document with the specified search target structure as the same group. For example, by setting [side effect] of a drug efficacy document as a target structure when extracting a characteristic character string and [effect] as a search target structure when calculating the similarity of a search target document,
It is possible to find a document that describes a drug that suppresses the side effects of the drug described in the seed document.

FIG. 6 is a diagram for explaining a problem of the second embodiment. In this example, [effect], [side effect], and [precautions] are specified as the structure relating to the seed document, and these structures are regarded as the search target structures, and the search is executed. As a result, feature strings that are less important or meaningless for the drug efficacy book such as “dose”, “car”, and “driving” are extracted, and as a result of similarity calculation based on the feature strings including these feature strings, , Document 2 and document 3 have such a value that the degree of similarity cannot be ignored, indicating that the search results include document 2 and document 3.

FIG. 7 is a diagram showing the configuration of the similar document search program 131b according to the third embodiment. In the third embodiment, a structure weighting program 600 is further added to the characteristic character string extraction program 150b.

FIG. 8 is a diagram showing a processing procedure of the third embodiment. The third embodiment is the same as the process of the second embodiment except that the structure weighting program 600 is applied to the characteristic character string extracted by the characteristic character string extraction program 150a. The structure weighting program 600 refers to a system definition file in which importance is set for each logical structure, and employs, for each logical structure, a characteristic character string extracted according to the importance for search. Determine the number of strings. For example, the structure of "important" employs all the extracted characteristic character strings, and the structure of "normal" employs only some characteristic character strings according to the importance of the extracted characteristic character strings. Alternatively, the number of characteristic character strings to be employed for each logical structure may be set in the system definition file, and a predetermined number of characteristic character strings having higher importance than the extracted characteristic character strings may be employed. Alternatively, a predetermined number may be adopted from a predetermined character type. As a method for calculating the importance of the characteristic character string, for example, Japanese Patent Application Laid-Open No. H11-143902 discloses a formula for calculating the importance of the characteristic character string as Expression 2. Instead of setting the importance of each logical structure and the number of adopted characteristic character strings in the system definition file, the importance may be specified as a part of the search condition expression via the input device 101. Note that the method of determining the characteristic character string to be adopted based on the importance of the logical structure or the importance of the characteristic character string is also applicable to the first embodiment.

By applying the similarity calculation program 152a and the corresponding structure determination program 401 after narrowing down the characteristic character strings as described above, documents with low importance can be excluded from the search results. Further, since the number of characteristic character strings is reduced as compared with the first and second embodiments, the search time for searching the appearance frequency file 181 can be reduced.

FIG. 9 is a diagram showing the configuration of the search result display program 132a according to the fourth embodiment. In the fourth embodiment, a structure-specific display method acquisition program 700 is added to the search result display program 132a.

FIG. 10 is a diagram showing a processing procedure according to the fourth embodiment. The fourth embodiment is the same as the processing of the first to third embodiments except for the processing of the search result display program 132a. Structure-based display method acquisition program 700
Displays the similarity of each document as a processing result of the similarity calculation program 152a for each search target structure. Also, the characteristic character strings extracted for each search target structure are highlighted.

FIG. 11 is a PAD diagram showing a processing procedure of the structure-specific display method acquisition program 700. The structure-based display method acquisition program 700 stores the characteristic character strings for each structure extracted by the characteristic character string extraction program 150a in the work area 161 (step 70).
1). Next, the similarity for each structure calculated by the similarity calculation program 152a is stored in the work area 161 (step 702). Next, the following processing is repeated for all the specified structures of the retrieved documents (step 703). First, the similarity of the structure stored in the work area 161 is obtained and displayed (step 704).
Next, the characteristic character string of the structure stored in the work area 161 is obtained and highlighted (step 705). In this embodiment, the similarity and the characteristic character string are highlighted for each logical structure, but only one of them may be performed. The display condition of the search result may be set in the system definition file or may be specified as a part of the search condition expression.

For the seed document and similar documents listed as a search result, a heading, an outline, etc. are displayed corresponding to the document numbers, and the characteristic character strings included and adopted in these texts are highlighted. Is also good. With such display, the characteristic character string included in the similar document can be compared with the characteristic character string included in the seed document.

Note that the appearance frequency file 181 used in the above embodiment may be replaced with an n-gram index disclosed in Japanese Patent Laid-Open No. 11-143902. That is, JP-A-11-143902 and JP-A-10-2407
By combining the structured character string index of Japanese Patent Publication No. 52, it is possible to configure a file that replaces the appearance frequency file 181. One or more occurrence positions of the same character string also indicate the number of appearances of the character string.

In the first to fourth embodiments, one document is designated as a seed document. However, a plurality of seed documents may be designated. Here, as the characteristic character string, all the characteristic character strings extracted from each seed document may be used, or a characteristic character string commonly included in each seed document may be used.

[0045]

As described above, according to the present invention, the designation of a logical structure is added as a search condition for a similar document search, so that the search accuracy can be improved while maximizing the advantage of the similar document search. . When a plurality of search target structures are designated, a less relevant feature character string can be excluded according to a predetermined importance of the logical structure or the importance of the characteristic character string, and the search accuracy can be further improved. . The seed document can be a structured document or an unstructured document, and the user does not need to pay attention to the selection of the seed document.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a similar document search system according to an embodiment.

FIG. 2 is a diagram illustrating a processing procedure according to the first embodiment.

FIG. 3 is a diagram showing a processing procedure of a conventional similar document search system.

FIG. 4 is a similar document search program 13 according to the second embodiment.
It is a figure which shows the structure of 1a.

FIG. 5 is a diagram illustrating a processing procedure according to a second embodiment.

FIG. 6 is a diagram illustrating a problem of the second embodiment.

FIG. 7 shows a similar document search program 13 according to the third embodiment.
It is a figure which shows the structure of 1b.

FIG. 8 is a diagram illustrating a processing procedure according to a third embodiment.

FIG. 9 shows a search result output program 13 according to the fourth embodiment.
It is a figure which shows the structure of 2a.

FIG. 10 is a diagram illustrating a processing procedure according to a fourth embodiment.

FIG. 11 is a diagram illustrating a processing procedure of a structure-specific display method acquisition program according to the fourth embodiment.

FIG. 12 is a diagram showing an example of a search condition input screen.

[Explanation of symbols]

131: similar document search program, 132: search result output program, 150: characteristic character string extraction program, 1
51: search target structure ID acquisition program, 152: similarity calculation program, 180: text, 181: appearance frequency file, 182: structure index, 400: seed document structure analysis program

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Natsuko Sugaya 890 Kashimada, Saiwai-ku, Kawasaki-shi, Kanagawa Prefecture Inside the Hitachi, Ltd.System Development Division (72) Inventor Yasuhiko Inaba 890 Kashimada, Saiwai-ku, Kawasaki-shi, Kanagawa Hitachi, Ltd. (72) Inventor Akihiko Yamaguchi 890 Kashimada, Saiwai-ku, Kawasaki-shi, Kanagawa Prefecture Hitachi, Ltd.System Development Headquarters (72) Yosuke Hachiji 5030 Totsukacho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Co., Ltd. F-term in Hitachi Software Division (reference) 5B009 QA09 VA02 5B075 ND03 NK06 NK39 PP13 PQ02 PQ22 PQ36 PQ46 PQ75 PR06 QM08

Claims (22)

[Claims] 1. A method for using a computer to search for a structured document similar to a document or a sentence (hereinafter collectively referred to as a seed document) specified as a search condition, wherein the seed is used as a search condition for similarity calculation. Receiving a designation of a document and at least one structure belonging to the structured document; and, after calculating the similarity, displaying a target document having a higher similarity with priority. Search method for similar documents. 2. A method for retrieving a structured document similar to a seed document using a computer, wherein when a seed document and a structure to be searched are specified, a feature is determined from the text of the specified seed document. Extracting a character string, and calculating a degree of similarity with the seed document based on the characteristic character string for a document in which the extracted characteristic character string matches the specified search target structure. Determining the display priority according to the order of the calculated similarity. The method for searching for similar documents by designating a structure. 3. A method for retrieving a document similar to a structured seed document using a computer, comprising the steps of receiving designation of a seed document and at least one structure belonging to the seed document; And displaying the target document having a higher degree of similarity with priority, after that. 4. A method for retrieving a document similar to a structured seed document using a computer, wherein when a seed document and a structure to be searched are specified, a text of the specified seed document is specified. Extracting a character string that is a characteristic from the text belonging to the specified structure among the texts, and extracting the character string based on the characteristic character string with respect to a document in which the extracted characteristic character string matches the specified search target structure. A method for searching for a similar document by designating a structure, comprising: calculating a similarity to a document; and determining a display priority in accordance with an order of the calculated similarity. 5. The method according to claim 1, wherein the seed document is a designated text on a display screen. 6. The similar document according to claim 1, wherein a structure to be excluded from the search target is specified instead of the structure to be searched. Search method. 7. The structure according to claim 3, wherein a structure belonging to the seed document and to be excluded from the search target is designated instead of a structure belonging to the seed document.
Search method for similar documents by specifying the structure of description. 8. When a plurality of structures are specified as search targets, a similarity is calculated for each of the search target structures, and a cumulative value of the similarities over all the search target structures is calculated as a final value at the time of determining the priority. The method according to claim 2 or 4, wherein the similarity is set to a similarity. 9. When a plurality of structures are specified as a search target, a similarity is calculated for each search target structure, and the highest similarity for the target document is determined as the final similarity at the time of determining the priority. The method according to claim 2 or 4, wherein a similar document is searched by designating a structure. 10. The method according to claim 2, wherein after extracting the characteristic character string, a characteristic character string to be adopted is further determined according to the importance of the characteristic character string. Search method for similar documents. 11. The method according to claim 4, further comprising, after extracting the characteristic character string, determining a characteristic character string to be adopted in accordance with the importance of the structure belonging to the specified seed document.
Search method for similar documents by specifying the structure of description. 12. The method according to claim 2, wherein after extracting the characteristic character string, a characteristic character string to be adopted is further determined according to the importance of the search target structure. Search method for similar documents. 13. The structure specification according to claim 2, wherein, when displaying the search result, the characteristic character string extracted for each of the search target structures is highlighted in the document to be displayed. Search method for similar documents. 14. A method for searching for a similar document by designating a structure according to claim 2, wherein, when displaying the search result, the similarity is displayed for each of the structures to be searched for the document to be displayed. . 15. An apparatus for searching a structured document similar to a document or a sentence (seed document) specified as a search condition, wherein the seed document and at least one document belonging to the structured document are used as search conditions for similarity calculation. An apparatus for retrieving similar documents by structure specification, comprising: means for receiving designation of two structures, and means for preferentially displaying a target document having a higher similarity after calculating the similarity. 16. An apparatus for retrieving a structured document similar to a seed document, wherein when a seed document and a structure to be searched are specified, a character string that is a feature is extracted from the text of the specified seed document. Means for extracting, for a document in which the extracted characteristic character string matches the specified search target structure, means for calculating the degree of similarity with the seed document based on the characteristic character string; Means for deciding display priority according to the order of the degree of similarity. 17. An apparatus for retrieving a document similar to a structured seed document, comprising: means for receiving designation of a seed document and at least one structure belonging to the seed document; Means for preferentially displaying a target document having a higher number of documents. 18. An apparatus for retrieving a document similar to a structured seed document, wherein when a seed document and a structure to be searched are specified, a specified one of the texts of the specified seed document is specified. Means for extracting a character string that is a feature from text belonging to the structure, and a similarity between the extracted feature string and a specified document to be searched that matches the seed document based on the feature string. And a means for determining the display priority in accordance with the order of the calculated similarity. 19. A program stored in a computer-readable storage medium, the program searching for a structured document similar to a document or a sentence (seed document) specified as a search condition, Program means for receiving designation of a seed document and at least one structure belonging to the structured document as search conditions for similarity calculation; and program means for giving priority to a target document having higher similarity after similarity calculation. A storage medium for storing a program, characterized by comprising: 20. A program stored in a computer-readable storage medium, the program searching for a structured document similar to a seed document, wherein a seed document and a structure to be searched are specified. Program means for extracting a character string that is a feature from the text of the specified seed document, and applying the feature character string to a document in which the extracted feature string matches the specified search target structure. A storage medium for storing a program, comprising: program means for calculating a similarity to the seed document based on the program; and program means for determining a display priority according to the order of the calculated similarity. 21. A program stored in a computer-readable storage medium, wherein the program is a program for searching for a document similar to a structured seed document,
A program characterized by comprising: program means for receiving a designation of a seed document and at least one structure belonging to the seed document; and program means for preferentially displaying a target document having a higher similarity after calculating the similarity. Storage medium for storing. 22. A program stored in a computer-readable storage medium, wherein the program is a program for searching for a document similar to a structured seed document,
When a seed document and a structure to be searched are specified, a program means for extracting a character string which is a feature from text belonging to the specified structure among texts of the specified seed document, and an extracted characteristic character string Program means for calculating a similarity with the seed document based on the characteristic character string for a document in which the specified search target structure matches, and determining a display priority in accordance with the calculated similarity in descending order. A storage medium for storing a program, comprising: