JP2000112953A - Literature retrieval method and its system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve a reproduction rate while the precision of retrieval is highly maintained. SOLUTION: A computer 1 reads patent literature (primary information) from a data base 3, extracts secondary information on a bibliographical item and an index word and stores them in an auxiliary storage device 7. A user inputs a retrieval expression by using a keyboard 4 and a pointing device 5 while he views a display device 2. When the method of the coappearance of the index word is regulated by an OR approximate operator in the retrieval expression, the computer 1 refers the inputted index word with the index word extracted from literature with a sentence range designated by the operator as a reference unit. Literature from which matched index words among the prescribed number of the index words, which is decided to originally be matched, and the index words which the user designates to be matched are extracted is selected as fitted literature.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for searching a large number of documents for a compatible document in which information desired by a user is described.

[0002]

2. Description of the Related Art Today, the importance of information has been greatly increased due to the arrival of the information society. In order to carry out various activities, obtain appropriate information and use it effectively,
And it has to be used efficiently.

[0003] For example, technical literature is essential information from the viewpoint of technology trend survey and shortening of R & D period.
However, as is clear from the fact that 500,000 to 600,000 patent documents (public publications and publications) are issued annually in Japan alone, the volume of technical literature is extremely large.
The patent document is very important information not only because it discloses the latest technical information, but also from the viewpoint of preventing patent litigation cases, which are becoming larger and more frequent. For this reason, the importance of information (document) retrieval for selecting important information from a vast amount of information is also increasing very much.

[0004]

A document search is performed by extracting all index words in a search formula input by a user (information requester, operator, or search agent) from documents from which index words (keywords) are extracted. Is performed by searching (selecting) documents that exist in the specified logical relationship as relevant documents.

[0005] The search formula is usually created by deciding the index words to be used and defining the relationship between the decided index words using an operator. The types of the operators can be roughly classified into a logical (Boolean) operator, a proximity operator, a comparison operator, and the like. Here, the logical operator is an operator that defines a co-occurrence relationship between a plurality of index terms, and the proximity operator is an operator that defines a manner in which index terms co-occur. The last comparison operator is an operator for specifying the range of the index word, and is used when it is desired to limit the range of the search.

[0006] The proximity operator is an operator introduced based on the premise that there is a conceptual connection between a plurality of index words that are located close to each other. The range of the place where it appears is specified. By using the proximity operator in a search expression to define the range in which multiple index terms co-occur, their order, etc.,
It is possible to improve the accuracy of the search.

However, an index word indirectly expresses a certain concept, and the correspondence between the concept and the index word is usually many-to-many. It is not usually possible to uniquely identify a concept from an index term. The proximity operator introduced in the conventional document search (system) basically only defines the range of places where a plurality of (two) index words appear together. For this reason, when the proximity operator is used in a search expression, the search accuracy is improved, while the search range of a plurality of index words is synergistically narrowed, and a search for a relevant document useful for the user is performed. There has been a problem that leakage is increased, that is, recall is reduced.

An object of the present invention is to improve the recall while maintaining high search accuracy.

[0009]

SUMMARY OF THE INVENTION A document retrieval method according to the present invention is based on the premise that it is used to retrieve a matching document by matching an index word input by a user with an index word extracted from the document. When the user specifies a plurality of index words to be matched only in a predetermined text range as search conditions, the plurality of index words are extracted from the text range extracted from the document. Then, a document in which an index term that matches at least an index term specified from a predetermined rule among a plurality of index terms by matching between index terms is extracted is selected as at least a matching document.

[0010] In the above method, the sentence range is:
It is desirable that the range is determined by allowing the user to select from a plurality of options prepared in advance. Also,
As a rule, the user may specify an index term that should at least match an index term extracted from a document among a plurality of index terms, and / or It is desirable to have the user specify an index word to be matched including the positional relationship. As the positional relationship, it is desirable to allow the user to specify the order in which two or more index words appear in the text range. Desirably, the options include at least one sentence and paragraph.

The document search system of the present invention is based on the premise that a matching document is searched by matching an index word input by a user with an index word extracted from the document, and the matching is limited to a predetermined sentence range. Search condition input means by which the user can input a plurality of index words to be searched as search conditions;
When the user inputs a plurality of index words to be limited to the sentence range by the search condition input unit, the matching unit matches the plurality of index words with the index words appearing in the sentence range extracted from the document. , Among a plurality of index words that are limited to the sentence range input by the search condition input means,
The document from which the index term collated by the collation unit is extracted with at least the index term specified from the predetermined rule is extracted,
Selecting means for selecting at least a relevant document.

[0012] In the above configuration, the sentence range is:
It is desirable to allow the user to select from a plurality of options prepared in advance by the search condition input unit. Further, as a rule, the search condition input means allows the user to specify an index word that should at least match an index word extracted from a document among a plurality of index words, and / or It is desirable to have the user specify an index word to be matched including the positional relationship with the index word extracted from the search condition using the search condition input means. As the positional relationship,
It is desirable to allow the user to specify the order in which two or more index terms appear within the text range. The above options are
It is desirable to include at least one sentence and paragraph.

The recording medium of the present invention is premised on that it is readable by a computer constituting a system for retrieving a compatible document by matching an index word input by a user with an index word extracted from the document. When the user inputs a plurality of index words to be matched only in the defined sentence range as a search condition, and the user inputs a plurality of index words to be matched only in the sentence range by the input means, Means for matching the plurality of index words with an index word appearing in a sentence range extracted from the document, and among a plurality of index words matched only for the sentence range input by means capable of inputting, A method of selecting, as at least a conforming document, a document from which an index term collated by a means for collating with at least an index term specified by a predetermined rule is extracted. When records the program for realizing.

In the document search method and system according to the present invention, when a user specifies a plurality of index words to be collated with a limited sentence range, the plurality of index words are extracted from the document. A document in which an index word that matches at least an index word specified by a predetermined rule among a plurality of index words is extracted is selected as a matching document at least.

When a plurality of index words are used as a search condition, the conceptual connection between the index words can be more accurately expressed (designated). On the other hand, when a rule is defined and an index word that must match among a plurality of index words is specified based on the rule, an allowable range is set for a condition that satisfies the rule using the plurality of index words. The degree to which the difference in the expressed index words affects the search result is reduced, and the search omission is reduced. From these facts, it is possible to improve the recall while maintaining high search accuracy.

[0016]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a configuration diagram of a document search system according to the present embodiment.

The system is for searching for patent documents, and is constructed by connecting various peripheral devices to a computer 1 for controlling the entire system. Specifically, for example, a display device 2 such as a CRT or an LCD, a database 3 having a primary information database described later, and an operator or a user (information requester, operator, search agent, etc.) transmit various data and commands, etc. A keyboard 4 for inputting, a pointing device (hereinafter referred to as a mouse as a representative thereof) 5, a printer 6 for forming an image on recording paper, and an auxiliary storage device (external storage device) such as an external hard disk device
7 is connected to the computer 1.

Although the system shown in FIG. 1 is constructed for searching patent documents using one computer, it is needless to say that a plurality of networks are used by using some kind of network such as LAN and public network. One system may be constructed by connecting computers. For the sake of convenience, unless otherwise specified, the patent documents are assumed to be published by the Japanese Patent Office for the purpose of disclosure, and will be used in the sense that they also include published and published gazettes of utility models.

[0019] These publications are now commonly referred to as CD-ROs.
M and is issued. Accordingly, the following description is based on the assumption that the CD-ROM 3b on which the publication is recorded and the CD-ROM autochanger (hereinafter abbreviated as "autochanger") 3a on which a plurality of the publications can be mounted are hardware corresponding to the database 3.

FIG. 2 is a functional block diagram of the above system. The figure shows a functional configuration by extracting functions particularly related to document search. The configuration and operation of the document search system according to the present embodiment will be described with reference to FIG.

As described above, the autochanger 3a shown in FIG. 1 has a CD-ROM 3b on which a patent document is recorded.
Is mounted. The patent document is primary information (primary material) because, for example, in Japan, it is a gazette issued by the JPO for the purpose of disclosure. For this reason, the autochanger 3
CD-ROM mounted on a (data recorded on) 3
b is the primary information database (fact database)
Corresponds to 11.

The medium for constructing the primary information database or storing the data is CD-RO.
It is not limited to M3b. Magneto-optical disk or D
Other types of media such as VD or magnetic tape may be used, and a plurality of types of media may be different. A database of primary information provided by another company or the like may be made accessible via some network (LAN, public network, or the like).

The target range designating section 12 is for an operator (system administrator or the like) to specify a target range for extracting a keyword from the patent document. The range is specified separately for the range of the patent document and the range in the patent document. Specifically, the scope of a patent document is specified by the filing date (or publication date), international patent classification, applicant, etc. The range within the patent document is, for example, the full text, only the claims (claims), And only the abstract sentence is selected by selecting one of the three. The designated results are sent to the keyword extracting unit 13. 1, the computer 1, the display device 2, the keyboard 4, and the mouse 5 correspond to the target range designation unit 12.

The keyword extracting unit 13 performs the primary information database 1 according to the specification result sent from the target range specifying unit 12, ie, the information indicating the target range specified by the operator.
Index words (keywords) for each patent document stored in
And stores it in the secondary information database 14. The computer 1, the autochanger 3a, and the auxiliary storage device 7 in FIG. 1 correspond to the keyword extraction unit 13 that extracts keywords in this way. Since the primary information database 11 is provided on the CD-ROM 3b, the keyword extraction unit 13
Is usually performed in units of the CD-ROM 3b.

The auxiliary recording device 7 is mainly used for recording the secondary information database 14. Therefore, the storage medium (data stored in) the auxiliary recording device 7 or provided therein corresponds to the secondary information database 14.

The secondary information database is a CD-RO
Secondary information used to search for patent documents (primary information) recorded in M3b is stored as data. As data of the secondary information, bibliographic items such as an application number, a publication number, an application date, a publication date, an applicant, an inventor, and an international patent classification are stored in addition to the keyword group extracted from the patent document. You. Relevant documents are searched using these data as clues.

The search condition input unit 15 is used by a user (information requester, search agent, or the like) to input search conditions (a search formula or the like) for searching for a suitable document with reference to the secondary information database 14. belongs to. 1, a computer 1, a display device 2, a keyboard 4, and a mouse 5 are shown.
Corresponds.

In the present embodiment, the following operators are introduced in addition to the various operators introduced in the conventional document retrieval system. Here, the operator is called a logical sum (OR) proximity operator for convenience.

The logical OR proximity operator is, for example, “(F)
A'.B. <CD >>. A to D are terms (keywords), "." Is an identifier by which the system recognizes a range defining a relationship between terms by a logical disjunction operator, and "F" in parentheses is those of them. Terms A ~
This is a symbol for designating a text range in which D appears. The “F” represents a paragraph. Text range symbols include:
In addition to “F”, “S”, “W”, “C”, and “A” are prepared. "S" is a sentence (sentence) from the period to the next period, and "W" is the number of words specified by the number added before it (one word if no number is added (two words are adjacent to each other). And "10W" is interpreted as 10 words), "C" is the text described in the claims, and "A" is the text described in the abstract. Respectively.

"'" Is a symbol (indispensable symbol) indicating that the term to which it is added must exist (match). "<" And ">" are symbols used to specify that a plurality of terms enclosed by them must exist (match) in the order described therein (order designation symbols) It is.

In the case where these symbols (options) are not used, in the present embodiment, it is sufficient that at least one term exists (coincides). In other words, it is assumed that the condition of co-occurrence of a plurality of terms defined by the logical sum proximity operator is satisfied. Those symbols (optional)
If no is used, it is stated that the number of terms that have been added by the number of terms specified by the user to be matched by those symbols must match. Therefore, for example, if one essential symbol is used, at least 2 (=
1 + 1) two terms must match.

Since the correspondence between a concept and an index word is usually many-to-many, it is assumed that an index word determined to express a certain concept is actually described in a patent document which is a relevant document. Not necessarily. For this reason, as in the related art, if the manner of co-occurrence of a plurality of index words is simply fixed, there is a high possibility that search omissions will increase. In order to avoid this, conventionally the number of index words handled by the proximity operator is set to two, but two index words cannot usually express much conceptual connection between them. For example, even if it is stipulated that “computer” and “design” appear in the same sentence (sentence), a patent that describes a completely different concept such as “computer design” and “computer design” is described. Search for references. This means that search noise is likely to occur. As is evident from this, it has not been possible to substantially increase the accuracy in the past.

On the other hand, by not limiting the number of index words to two as in the present embodiment, conceptual connections between index words can be specified more accurately. Therefore, the expressive power of the concept is improved. For example, if an object to be designed by computer is added as an index word, it is possible not to search for documents in which the text "computer design" is simply described. This means that the accuracy can be easily improved.

On the other hand, index words that must appear in the specified text range can be specified, so that index words with high discrimination can be used effectively. On the other hand, it is possible to specify an index word that does not necessarily appear in the specified sentence range. In other words, while providing an allowable range for the condition that satisfies the rule using a plurality of index words, Since the allowable range can be changed, it is possible to reduce search omission due to a difference between an index term expressing a concept (an index term extracted from a document and an index term described by a user in a search formula). You can do it.

For example, in order to search a document describing a method for preventing double conveyance of a medium such as a recording paper, five index words of “medium, double, conveyance, prevention, mechanism” were selected. Document that describes the concept represented by the five index terms (actually relevant documents)
Therefore, these index words are not always used to express the concept. "Double, transport, prevention, mechanism"
Or "medium / double / transport / prevention / device" or "medium /
The concept may be expressed using a phrase such as “multiple / extending / avoidance / mechanism”. Documents using such phrases cannot be searched conventionally. However, in the present embodiment, since all index words do not necessarily have to match, even a document in which such a term is used can be searched as a matching document.

As is apparent from the above, in the present embodiment, it is possible to easily easily substantially improve the recall and accuracy. When the order designation symbol is used as necessary, the discrimination power is improved thereby, and the accuracy can be particularly improved.

In the present embodiment, at least one term exists among a plurality of terms (index terms) defining the manner of appearance by the logical disjunction operator, that is, a term extracted from a document and If they match, the specified condition of appearance is satisfied. However, the number of terms to be matched may be changed according to the number of terms specified by the user.
For example, the number of terms to be matched may be changed, such as one for up to three terms, two for four to six terms, and three for more. Further, the user may be able to specify the number of terms to be matched.

Further, symbols for specifying the order (“<” and “>”)
Is used to specify (prescribe) that the terms enclosed by them must exist in that order, but also specify the terms that must be adjacent regardless of order ( May be specified or / and / or the interval between terms that must be present may be specified (defined). If the word spacing can be specified, it or one of them may appear more than once in one sentence. In other words, there must be more than one corresponding word spacing. Therefore, the order in which terms appear may be specified together so that the accuracy can be further improved as necessary. As a matter of course, the positional relationship between other terms may be specified (defined).

In addition to terms that may appear in the designated text range, terms that are unlikely to appear together with them or that are considered to be extremely unlikely may be separately designated. As apparent from these, various modifications can be made.

As described above, the search condition input unit 15 can create a search expression (search condition) by using the logical OR proximity operator. The search condition is sent from the search condition input unit 15 to the search and extraction unit 16.

The search / extraction section 16 refers to the secondary information database 14 to search for a suitable document in which all index words specified by the search condition (search formula) sent from the search condition input section 15 exist. . Although various operators can be used in the search expression, if the above-mentioned logical OR proximity operator is used in the search expression, an index term specified to be matched by the ethical OR proximity operator And at least the extracted index terms must match (required, of course, may be sufficient)
Becomes For this reason, a document that satisfies at least the necessary conditions is selected as the relevant document. In FIG. 1, the computer 1 and the auxiliary storage device 7 correspond to the search and extraction unit 16 that searches for such a suitable document.

The index term stored in the secondary information database 14 is a free term because it is a word extracted from a document. In a free term, the difference in the concept that the same term expresses tends to increase. This means that search leakage is likely to occur. In order to avoid this, in the present embodiment, it is possible to perform a search using terms that are arranged and controlled.

The user can input a thesaurus (descriptor (control word) registered as an index word, a non-descriptor that is a synonym / semi-synonym, an upper word, a lower word, a related word,
A controlled term is input as an index word in the search condition input unit 15 by referring to a search vocabulary composed of a symbol indicating a type of relation between words, a scope note indicating a meaning range and usage of a headword, and the like. can do. For this reason, as one of the search conditions, the user is made to select whether or not to perform a search using arranged and controlled terms (hereinafter also referred to as controlled words). When a controlled word is selected, the input index word is treated as a controlled word.

When the user instructs a search using a controlled word via the search condition input unit 15, the search and extraction unit 16 converts the index word stored in the secondary information database 14 into a controlled word, or By referring to a synonym dictionary, a high-level word dictionary, or the like prepared according to the thesaurus to be referred to, collation between index terms is performed to search for a matching document. Thereby, the recall can be further improved.

Upon completion of the search performed as described above, the search extraction unit 16 displays the search result on the display device 2. If the search result is specified to be printed out via the search condition input unit 15, or
When printout is instructed after the search, the printer 6
To print out.

The search condition input section 15 is capable of instructing a printout of the patent document itself stored in the primary information database 11. When the printout is instructed by designating the searched patent document, or when the printout is instructed simply by designating the patent document, the search and extraction unit 16 retrieves the data of the designated patent document from the primary information database 11. The data is read out and output to the printer 6 for printing.

In the present embodiment, necessary information is searched for and provided to the user by executing the processing related to the document search as described above. Note that the target range specifying unit 12, the keyword extracting unit 13, the search condition input unit 15, and the
The function of each unit of the search and extraction unit 16 is realized when the CPU mounted on the computer 1 executes a program stored in a hard disk device (both not shown) built in the computer 1.

Next, the operation of the computer 1 for realizing the functions of the units 12, 13, 15, and 16 will be described in detail with reference to various flowcharts shown in FIGS.

FIG. 3 is a flowchart of the index term extraction process. This processing is executed to realize the functions of the target range specifying unit 12 and the keyword extracting unit 13 shown in FIG. 2 when, for example, the operator instructs the extraction of the index word via the keyboard 4 or the mouse 5. You.

First, in step S1, the keyboard 4,
Alternatively, the operator is caused to input, via the mouse 5, a target range for extracting an index word. Thereby, as described above,
The target range is designated by dividing the range into the patent document and the range in the patent document. In a succeeding step S2, it is determined whether or not the input has been completed, that is, whether or not extraction of the index word has been instructed.
If the operator has not issued the instruction, the determination is NO and the process returns to step S1. Otherwise, the determination is yes and the process moves to step S3.

In step S3, the data of the CD-ROM 3b (primary information database 14) corresponding to the target range designated by the operator is read out by the autochanger 3a, an index word is extracted for each document, and stored in the auxiliary storage device 7. That is, the index words are stored in the secondary information database 14.

The extraction of the above-mentioned index words is specifically performed as follows, for example. The hard disk device built in the computer 1 stores various dictionaries for text analysis such as an independent word dictionary, an auxiliary word dictionary, a synonym dictionary, a high-order word dictionary, syntax rules, and various tables. As described above, the synonym dictionary, the high-order word dictionary, and the like are dictionaries that are also referred to during document search.

The extraction of the index word is performed by extracting a term for each patent document with reference to the various dictionaries, specifying an important word from the extracted words, and using the specified important word as an index word. I have. Thereby, for example, two or more characters composed of a kanji character and a katakana character immediately before a word constituting a title, a particle which subjects a sentence, and a compound lexical character (“ha”, “ni”, “toha”, etc.) A phrase or a pre-defined importance determination rule (eg, a term immediately preceding a word such as “by,” “by,” “by,” “as,” “to” is considered to be more important to the sentence containing it The phrase corresponding to (rule) is specified as an important word (index word).
It is also important to examine the frequency of occurrence of the cut-out terms, and to find the terms that appear very frequently or have a relatively low frequency of appearance, in other words, terms that have a frequency of appearance between high-frequency words and low-frequency words. It is specified as a word (index word).

The index words extracted as described above are sentence range according to the location where they appear so that they can correspond to a search in which the relation between index words is defined by an operator such as a logical OR proximity operator. (Information indicating the positional relationship between index words may be separately generated). At this time, information indicating the number of words between terms (called inter-word information) is also acquired. When these steps are completed, steps S3 to S
Move to 4.

In step S4, the auxiliary storage device 7 simply stores the index words extracted from the document and the index word group (including the inter-word information in this case) compiled for each document unit as described above.
, Ie, they are stored in the secondary information database 14 as secondary information. After that, a series of processing ends.

Although not particularly shown in FIG. 3, bibliographic information such as the date of application, the date of publication, and the applicant is also made in accordance with the extraction of index words. Thus, after the execution of the index word extraction processing, all the secondary information on the patent documents in the range specified by the operator is stored in the secondary information database 14.

FIG. 4 is a flowchart of the search process. This process is executed to realize the functions of the search condition input unit 15 and the search extraction unit 16 shown in FIG. 2 when, for example, the user instructs a document search via the keyboard 4 or the mouse 5.

First, in step S11, the operator is caused to input search conditions (mainly search expressions) via the keyboard 4 or the mouse 5. At this time, the user can input a search expression using the above-mentioned OR proximity operator. In a succeeding step S12, it is determined whether or not the input has been completed, that is, whether or not a search execution has been instructed. If the user has not given the instruction, the determination is NO and the process returns to step S11. Otherwise, the decision is YE
In S, the process moves to step S13.

In step S13, the data of the secondary information database 14 is read from the auxiliary storage device 7, and for each document, index words are collated in accordance with the rules of the search condition (search formula) input by the user. Relevant documents in which all index words exist are extracted. At this time, if the proximity operator exists in the search expression, the matching term is collated in the text range specified by the proximity operator to extract a matching document. When the user instructs to collate the index word as a controlled word, the concept of the index word read from the secondary information database 14 and the upper and lower Relevant documents are extracted by collating between index terms in consideration of terms of concepts having similar or similar relationships. In this way, the collation with the patent documents in the range designated by the user is sequentially performed, and when all the matching documents based on the collation results in the range are extracted, the process proceeds to step S14.

In step S14, the result of the search performed as described above is output. Specifically, the search result is displayed on the display device 2 and, if instructed by the user in advance, the search result is printed out on the printer 6. After that, a series of processing ends.

In the present embodiment, the search is performed for patent documents, but the present invention is not limited to patent documents. Not only scientific literature typified by patent literature, but also literature related to economy and industry, newspaper articles, and the like can be searched.

A computer (system) having hardware necessary for performing a document search is loaded with a program for realizing the flowcharts shown in FIGS. 3 and 4 or at least the flowchart shown in FIG. Thus, the system as in the present embodiment can be realized. From this, the program can be stored on CD-R
The data may be recorded on a recording medium such as an OM, a DVD, a floppy disk, or a magneto-optical disk and distributed. Or,
Alternatively, distribution may be performed via some network (for example, a public network).

[0063]

As described above, according to the present invention, when a user specifies a plurality of index words to be collated with a limited sentence range as a search condition, the plurality of index words are extracted from the document extracted from the document. A document from which an index term that matches at least an index term specified by a predetermined rule among a plurality of index terms is extracted is selected as a matching document at least. For this reason, it is possible to improve the recall while maintaining high search accuracy.

[Brief description of the drawings]

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

FIG. 2 is a functional block diagram of the document search system according to the embodiment;

FIG. 3 is a flowchart of an index term extraction process.

FIG. 4 is a flowchart of a search process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Computer 2 Display device 3 Database 3a CD-ROM autochanger 3b CD-ROM 4 Keyboard 5 Pointing device 7 Auxiliary storage device 11 Primary information database 12 Target range specification part 13 Keyword extraction part 14 Secondary information database 15 Search condition input part 16 Search Extractor

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Jun Tarishi 1776 Yanoguchi, Inagi-shi, Tokyo Fujitsu Kiden Co., Ltd. (72) Inventor Hidehiko Nakamura 1776 Yanoguchi, Inagi-shi, Tokyo Fujitsu Kiden F-term (reference) 5B075 ND03 NK04 NK06 NK35 NK37 NK50 PP02 PP03 PP23 PQ02 PQ03 QS01 UU06

Claims (13)

[Claims] 1. A method for searching for a matching document by matching an index word input by a user with an index word extracted from a document, wherein the matching is performed only in a predetermined sentence range. When the user specifies a plurality of search terms as search conditions, the plurality of index terms are checked against index terms appearing in the sentence range extracted from the document, and the plurality of index terms are checked by matching between the index terms. Selecting a document from which an index word at least matching an index word specified by a predetermined rule is extracted as at least the relevant document. 2. The document search method according to claim 1, wherein the sentence range is a range determined by allowing a user to select from a plurality of options prepared in advance. 3. The method according to claim 1, wherein the rule is that a user designates an index term that should at least match an index term extracted from the document among the plurality of index terms. Literature search method. 4. The method according to claim 1, wherein, as the rule, a user designates an index word to be matched with the index word extracted from the document including a positional relationship among the plurality of index words. Document search method described in 2, 3 or 3. 5. The document search method according to claim 4, wherein, as the positional relationship, a user can specify an order in which two or more index words appear in the sentence range. 6. The document search method according to claim 2, wherein the options include at least one sentence and a paragraph. 7. A system for searching for a matching document by matching an index term input by a user with an index term extracted from a document, wherein the matching is performed only in a predetermined text range. A plurality of search condition input means capable of inputting a plurality of search terms as search conditions, and when the user inputs a plurality of index terms limited to the sentence range by the search condition input means, A matching means for matching with an index word appearing in the sentence range extracted from the sentence, and a predetermined rule among a plurality of index words matched only for the sentence range input by the search condition input means. And selecting means for selecting, as at least the relevant document, a document from which the index word collated by the collating means is at least matched with the index word specified from. Document retrieval system according to claim. 8. The document retrieval system according to claim 7, wherein the user is allowed to select the sentence range from a plurality of options prepared in advance by the search condition input unit. 9. The search condition input unit, as a rule, causes a user to specify an index word that should at least match an index word extracted from the document among the plurality of index words. Item 7 or 8, a document search system. 10. The method according to claim 1, wherein, as the rule, a user specifies an index word to be matched with an index word extracted from the document including a positional relationship among the plurality of index words by the search condition input unit. 10. The document search system according to claim 7, 8, or 9. 11. The document retrieval system according to claim 10, wherein, as the positional relationship, an order in which two or more index words appear in the sentence range can be specified by a user. 12. The document search system according to claim 8, wherein the options include at least one sentence and a paragraph. 13. A computer-readable recording medium constituting a system for searching a matching document by matching an index word inputted by a user with an index word extracted from a document, wherein the computer-readable recording medium has a predetermined text range. Means for allowing the user to input a plurality of index words to be collated in a limited manner as search conditions; and when the user inputs a plurality of index words to be collated and limited to the sentence range by the input means, the plurality of Means for matching an index word with an index word appearing in the sentence range extracted from the document; and among a plurality of index words matched only for the sentence range input by the input enabled means, The document from which the index term matched by the matching means is extracted at least to match the index term specified from the predetermined rule is defined as at least the matching document. A computer-readable recording medium on which a program for realizing the above is recorded.