JP2003281171A - Document retrieval device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document retrieval device, enabling retrieval of a document where a retrieval work appears regardless of whether or not the retrieval word of inflection type is normalized to a representative word form. <P>SOLUTION: When a registration processing part 12 stores information including an identifier of a document where a character string extracted by a text processing part 11 appears and the appearance position of the character string in document data stored in a document storage part 16 in association with the character string in an index storage part 17. When the character string is an inflection form, a representative word form to the character string is stored in association with the character string in the index storage part 17. A register processing part 12 stores the representative word form and a different character string from the representative word form in the inflection form in association with the same appearance position information in the index storage part 17. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval device for retrieving document data desired by a user from a document database.

[0002]

2. Description of the Related Art Conventionally, when a full-text search is performed on document data described in a language in which word delimiters are clearly defined, such as English, a character string to be registered in an index as an index word is cut out from the document data. In general, a character string before and after the blank or punctuation mark is generally used as a delimiter.

However, when an index is created in this way, a language having a suffix change such as the past tense or plural, such as English, eg, "eg document / document".
In "s, explore / explores / explored / exploring", even if a representative word form such as "explore" is specified as a search condition, it does not match the inflectional form in the document, and a search omission occurs. Or, as in English, inflected forms (verb past tense and noun plural forms (eg explore / explored, explorer / explorer
s)) and derived forms (noun forms, adjective forms, etc. (eg explore / exp)
In languages with inflected forms such as loration / exploratory)), even if a stem such as "explore" is specified as a search condition, it does not match the inflected forms in the document, resulting in omission of search.

In order to deal with such a problem, a process (stemming) for normalizing and registering an index word into a representative word form is generally performed at the time of index creation. Porter (1980) and Lovins (1968) are well known as conventional techniques for stemming. These are "Porter, MF (1980)" A
n algorithm for suffix stripping "(Program, 14 (3),
pp.130-37) "," Lovins, J. (1968) "Development of
a stemming algorithm "(Mechanical Translation and
Computational Linguistics, 11, pp.22-31) ”. Here, whether or not the representative word forms and the inflectional forms or the inflectional forms having a common representative word form may be identified with each other differs depending on the appearing document and the search condition. When normalized at the time of index registration, the representative word form and the inflectional form are always regarded as the same, and it becomes impossible to avoid unnecessary search results depending on the search condition.

Therefore, as a method of solving the problem, there is a method of expanding a search word into a related notation and searching at the time of search, instead of normalizing at the time of index registration. A document search device using this method is disclosed in Japanese Patent Laid-Open No. 7-65013. Here, the search terms are expanded by referring to a dictionary that stores different notations and synonyms.

[0006]

However, in such a conventional document search apparatus, since the search words are expanded by all the related notations and word forms, the number of expanded words becomes very large and the search efficiency is remarkably lowered. There was a problem of fear.

The present invention has been made in order to solve such a problem, and a document in which the search word appears regardless of whether or not the search word which is the inflectional form is normalized to the representative word form. A document search device capable of searching for a document is provided.

[0008]

According to a first aspect of the present invention, there is provided a document search apparatus, wherein a document is divided into words and is described in a language having a ending change. Document storing means for adding and storing, inquiry input means for inputting inquiry text for searching a desired document from the document data stored by the document storing means, and document data stored by the document storing means Alternatively, a character string information extracting unit that extracts character string information including a character string delimited by a predetermined delimiter as a boundary, and character string information including an appearance position of the character string, for the inquiry text input by the inquiry input unit, A sentence in which the character string extracted by the character string information extracting unit appears in the document data stored by the document storing unit The information including the identifier and the appearance position of the character string is stored in the index in association with the character string, and, when the character string is inflection, the representative word form for the character string is the character string. A search condition of a predetermined format is created based on the character string extracted by the character string information extracting unit from the index storage unit that stores the data in association with the index and the inquiry text input by the inquiry input unit. It has a structure provided with a search condition creating means and a search processing means for referring to an index stored by the index storage means to search for a document satisfying the search condition created by the search condition creating means. . With this configuration, it is possible to search for a document in which the search word appears, whether the search word that is the inflected form is normalized to the representative word form or not.

A document retrieval apparatus according to claim 2 of the present invention is
2. The index storage means according to claim 1, wherein the character string extracted by the character string information extracting means is an inflected form, and the inflected form completely includes a representative word form for the character string. The representative word form and the character string different from the representative word form in the inflectional form are stored in the index in association with the same appearance position information.
With this configuration, it is possible to avoid double-registering one character string in the document data as an index and suppress an increase in index size.

A document search device according to claim 3 of the present invention is
In Claim 1, when the character string extracted from said inquiry text by said character string information extraction means is an inflectional form, said inflectional form and the representative word form in said document are the same by said search processing means. Designation means for designating whether or not to search by considering it as a word, or whether or not to search by considering the representative word form of the character string extracted from the inquiry text and the inflection form in the document as the same word It has the configuration provided. With this configuration, it is possible to perform a flexible search according to the designation of the user.

A document retrieval apparatus according to claim 4 of the present invention is
From the document data stored by the document storage means for storing the electronically-documented document data, which is delimited for each word and described in a language having a change in word, by giving an identifier for each document Inquiry input means for inputting an inquiry text for searching a desired document, and the document data stored by the document storage means or the inquiry text input by the inquiry input means, with a predetermined delimiter as a boundary Character string information extracting means for extracting character string information including the delimited character string and the appearance position of this character string, and document data stored by the document storing means, extracted by the character string information extracting means. The information including the identifier of the document in which the character string appears and the appearance position of the character string are associated with the character string. When the character string stored in the index and extracted from the document data by the character string information extracting means is the inflection, the stem character obtained by decomposing the character string into a stem and a stem portion. Index storage means for storing a string in the index in association with the same appearance position information as the character string, and a character string extracted by the character string information extracting means from the inquiry text input by the inquiry input means. A search condition creating means for creating a search condition in a predetermined format and a document stored in the index storage means are searched for a document satisfying the search condition created by the search condition creating means. Search processing means for performing the search processing. With this configuration, it is possible to search for a document in which the search word appears, whether or not the inflection form that is the search word is normalized to the word stem.

A document retrieval apparatus according to claim 5 of the present invention is
5. The index storage means according to claim 4, wherein when the character string extracted from the document data by the character string information extracting means is a word inflection, the character string is decomposed into a word stem and a word end portion. The stem string and ending character string
The index is stored in the index in association with the same appearance position information. With this configuration, it is possible to avoid double registration of the common part of the inflectional form and the stem, and to suppress the increase in the index size.

A document retrieval apparatus according to claim 6 of the present invention is
6. The index storage means according to claim 5, wherein when the ending character string obtained by decomposing the character string into a stem and a ending portion is composed of a plurality of consecutive endings, the ending character string is further decomposed. The plurality of ending character strings and the stem character strings obtained in this way are stored in the index in association with the same appearance position information. With this configuration, it is possible to avoid double registration of the common part of the inflection and the stem, suppress the increase of the index size, and at the same time, enable more flexible search.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] FIG. 1 is a block diagram showing a hardware configuration of essential parts of a document search apparatus according to a first embodiment of the present invention. In FIG. 1, a document search device 1 includes a central processing unit (hereinafter also referred to as a CPU) 2 having a function of controlling each unit, a memory 3 for storing a control program of the CPU 2, and the like. A hard disk 4 for storing documents to be searched and attribute data, an input unit 5 for giving search conditions and various instructions using a keyboard, a mouse, etc.
A display unit 6 including a CRT or a liquid crystal display (LCD) for displaying search results and the like, and a flexible disk drive (hereinafter also referred to as FDD) for reading / writing data from / to a flexible disk (hereinafter also referred to as FD).
7. CD-ROM drive 8 for reading data from CDROM (hereinafter also referred to as CD-ROM), controlling connection with a communication line via a communication interface, and communicating with other communication devices through this communication line Also, a communication unit 20 for exchanging data and the like are connected by a bus 9.

FIG. 2 is a block diagram showing the functional arrangement of the essential parts of the document searching apparatus according to the first embodiment of the present invention. In FIG. 2, the memory 3 is provided with an input processing unit 10, a text processing unit 11, a registration processing unit 12, a search condition creation unit 13, a search processing unit 14, and an output processing unit 15, and the execution is controlled by the CPU 2. . Further, the hard disk 4 includes a document storage unit (hereinafter also referred to as a document DB) 16 that stores a large number of document data as a document database (document DB), an index storage unit 17 that stores an index of each document data, A data storage unit 18 for storing the ending normalization data used in the text processing unit 11 is formed.

The hard disk 4, which is a recording medium,
Various programs such as a document search program are stored in the FD or the CD-ROM, and are written in the memory 3 when the system of the document search device 1 is started.

First, the outline of the index registration process in such a configuration will be described. The input processing unit 10 sends the registration instruction input from the input unit 5 to the registration processing unit 12. Registration processing department
12 is the document DB according to the instruction sent from the input processing unit 10.
Document data is read from 16 documents one by one, and the read document data is sent to the text processing unit 11. The text processing unit 11 processes the sent data to generate an index word (character string).
Is extracted and returned to the registration processing unit 12 together with the appearance position information. Further, the text processing unit 11 reads the word end normalization data stored in the data storage unit 18 into the memory 3 in accordance with the normalization processing parameter designated at the time of system activation, and performs the normalization processing.

Next, the outline of the search process will be described. The input processing unit 10 sends the search inquiry input from the input unit 5 to the search condition creating unit 13. The search condition creation unit 13 sends the inquiry text to the text processing unit 11. The text processing unit 11 processes the sent inquiry text in the same manner as at the time of index registration, extracts a search word (character string), and sends it to the search condition creation unit 13. Further, the text processing unit 11 refers to the parameter of the normalization process designated at the time of index registration, and correspondingly reads the end-of-word normalization data stored in the data storage unit 18 into the memory 3 to perform the normalization process. , To the search condition creation unit 13. The search condition creation unit 13 converts the sent search word into an operator format, and creates search conditions that can be processed by the search processing unit 14. The search processing unit 14 searches the index stored in the index storage unit 17 according to the search condition, and identifies the document that satisfies the condition. Output processing unit
At 15, the information of the document based on the index searched by the search processing unit 14 is displayed on the display unit 6. If necessary, the document DB 16 is accessed to output the document data.

Next, the document registration process and the document search process executed by the CPU 2 according to the document search program written in the memory 3 as described above will be described. First, the document registration process will be described. Here, FIG. 3 is a flowchart schematically showing the flow of document registration processing in the registration processing unit 12, and FIG. 4 is a flowchart schematically showing the flow of character string information extraction processing at the time of registration in the text processing unit 11.

As shown in FIG. 3, in the document registration process, first, it is determined whether or not the document data stored in the document DB of the hard disk 4 includes a document for which an index has not been created (step S1). . If there is no document for which an index has not been created (N in step S1), indexes of all document data to be searched have been created.
The document registration process ends.

On the other hand, when there is a document for which an index has not been created (Y in step S1), the document for which an index has not been created is read (step S2), and a document ID uniquely indicating each document stored in the document DB 16 is acquired. (Step S
3), character string information extraction processing for document data (shown in FIG. 4)
To (step S4). If it is an end instruction (Y in step S5), it means that the document being processed is finished.
Returning to step S1, the registration process is repeated. If it is not the end instruction (N in step S5), the process returns to step S4 to repeat the character string information extraction processing. The character string information returned from the character string information extraction process is registered as an index in the index storage unit 17 (step S6), and the process returns to step S4 to repeat the character string information extraction process.

As shown in FIG. 4, in the character string information extraction processing, the text processing section 11 first checks whether or not the registration mode is designated, and if not, performs the search mode processing (step S10). N). If the registration mode is designated (Y in step S10), the appearance position is set to zero (initialization) (step S11), and the start position is set (step S12). If the end of the document has been reached (Y in step S13), the end instruction is returned (step S1).
4), the process ends. If the end of the document has not been reached (N in step S13), the index word is extracted from the start position by removing the delimiters such as spaces and punctuation marks (step S15). Then, the parameter of the normalization process designated at the time of starting the system is checked (step S16). The parameter values will be described later. If the parameter value is 0 (Y in step S16), the extracted index word notation and appearance position are returned (step S17),
Advance the appearance position by one (step S18), step S12
Return to and repeat the process. If the parameter value is other than 0, a normalization process (step S19) described below is performed, and then the process proceeds to step S17 and subsequent steps.

Next, the document search process will be described. Here, FIG. 5 is a flowchart schematically showing the flow of the document search process from the search condition creation unit 13 to the output processing unit 15.
FIG. 6 is a flow chart schematically showing the flow of character string information extraction processing at the time of search in the text processing unit 11.

As shown in FIG. 5, in the document search process, first, the input processing unit 10 sends the inquiry text input from the input unit 5 to the search condition creating unit 13 (step S
20), the search condition creation unit 13 sends the sent inquiry text to the character string information extraction processing (shown in FIG. 6) (step S21). If there is no end instruction from the character string information extraction process (N in step S22), the search word returned by the character string information extraction process is stored (step S23), and the process returns to step S21 to repeat the process. When the end instruction is returned (step S
22 Y), a search condition is created and a search process is performed (step S24), and the search result is output (step S25).

As shown in FIG. 6, in the character string information extraction processing, the text processing unit 11 first checks whether or not the search mode is designated, and if not designated (N in step S30), the registration mode is set. Is processed. If the search mode is designated (Y in step S30), the start position is set (step S31), and if the end of the text is reached (Y in step S32), the end instruction is returned (step S).
33), and ends the process. If the end of the text has not been reached (N in step S32), the search word is extracted from the start position, excluding delimiters such as spaces and punctuation marks (step S34). Next, referring to the parameter of the normalization process designated at the time of index registration (step S35), if the parameter value is 0 (Y of step S35),
The extracted search term is returned (step S36) and step S31.
Return to and repeat the process. If the parameter value is other than 0, the normalization process (step S37) described later is performed, and then the process proceeds to step S36 and the subsequent steps.

Next, the normalization process will be described. In addition, in the present embodiment, the following four values are set as the parameters of the normalization process at the time of index registration. 0: Only denormalized word forms are registered in the index. 1: Only the normalized word form is registered in the index. 2: Register both the denormalized word form and the normalized word form in the index. 3: Register the ending parts of the normalized word form and the denormalized word form in the index.

In the present embodiment, whether the inflectional form and the representative word form are regarded as the same word at the time of searching (normalization mode), or not regarded as the same word (non-normalization mode). , Shall be specified. Here, the parameter and each mode may be designated by the user. The parameters and each mode designated by the user are input, for example, by the input unit 5 and are stored in the hard disk 4 or the memory 3 in an updatable manner. Correspondence between the parameter values at the time of registration and the feasibility of each mode at the time of search is as follows. 0: Only non-normalized mode is possible. 1: Only normalization mode is possible. 2: Both modes are possible. 3: Both modes are possible.

FIG. 7 is a description example of the ending normalization data. In this embodiment, dictionary format data for word notation is used. The dictionary data is composed of three fields, the first field is a headword form, and the inflection form to be normalized is described. In the second field, the representative word form for the headline is described. The third field is an embodiment of the invention according to claim 2, and is a tail field in which the difference character string of the tail part is described when the headword form completely includes the representative word form. For example, since the entry word form "accounting" completely includes the representative word form "account", the difference character string "ing" is described in the ending field. On the other hand, the headword form "classified" does not completely include the representative word form "classify", and thus the ending field is not described.

FIG. 8 is a flow chart showing an outline of the normalization processing at the time of registration and search when the value of the normalization parameter at the time of index registration is 1. The text processing unit 11 performs a dictionary search using the index word or the search word as a key (step S40), and if there is no matching heading (N in step S41), the process ends. If there is a matching headline (Y in step S41), the corresponding representative word form is returned (step S42), and the process ends.

FIG. 9 is a flow chart showing an outline of the normalization process at the time of registration when the value of the normalization parameter at the time of index registration is 2. The text processing unit 11 performs a dictionary search using the index word as a key (step S50),
If there is no matching headline (N in step S51), the process ends. If there is a matching headline (Y in step S51), both the headline word form and the representative word form are returned (step S52), and the process ends.

FIG. 10 is a flow chart showing an outline of the normalization process at the time of search when the value of the normalization parameter at the time of index registration is 2. If the parameter value is 2 or more, it is possible to specify the normalization mode or the non-normalization mode. Therefore, if the text processing unit 11 is not the normalization mode (N in step S60), the process ends. If it is in the normalization mode (Y in step S60), a dictionary search is performed using the search word as a key (step S61). If there is no matching headline (N in step S62), the process ends. If there is a matching headline (Y in step S62), the corresponding representative word form is returned (step S63), and the process ends.

FIG. 11 is a flow chart showing an outline of the normalization process at the time of registration when the value of the normalization parameter at the time of index registration is 3. The text processing unit 11 performs a dictionary search using the index word as a key (step S70),
If there is no matching headline (N in step S71), the process ends. If there is a matching headline (Y in step S71), both the headline word form and the representative word form are returned (step S72), and the process ends.

FIG. 12 is a flow chart showing an outline of the normalization process at the time of search when the value of the normalization parameter at the time of index registration is 3. The text processing unit 11 first performs a dictionary search using the search word as a key (step S
80), if there is no matching headline (N in step S81), the process ends. If there is a matching headline (Y in step S81), it is determined whether or not it is in the normalization mode. If it is in the normalization mode (Y in step S82), the representative word form is returned (step S83), and the process is ended. . If it is not the normalization mode (N in step S82), it is confirmed whether or not the ending field corresponding to the headword form is described. If it is not described (N in step S84), the process ends. If it is described (Y in step S84), both the representative word form and the ending are returned (step S85), and the process ends.

Next, a specific example of the document registration process according to this embodiment will be described. Here, FIG. 13 is a schematic diagram of a document to be registered, and in this embodiment, it is assumed that there are four documents A to D.

Document A contains "Classified and Shop" in the document.
The document data is an English document data including a character string "ping", and an identifier (document ID: 1) uniquely identifying the document A is added. The document B has "computer account reques" in the document.
Document data in English including a character string "t"
Is uniquely assigned (document ID: 2). The document C is English document data including a character string "accounting firm hired" in the document, and is given an identifier (document ID: 3) that uniquely indicates the document C. The document D is English document data containing a character string "classifying pop rock" in the document, and is given an identifier (document ID: 4) uniquely indicating the document D.

14 to 16 are schematic diagrams of indexes when the values of the normalization parameters at the time of registration are 1 to 3, respectively. This index is an index word notation,
It is composed of a document ID assigned to the document in which the index word appears, an appearance position in each document, and a search constraint flag indicating a constraint at the time of search. The search constraint flag is 0 when there is no constraint, 1 when the index word is the representative notation generated by the normalization process, and 1 when the index word is the ending notation generated by the normalization process. 2 is registered. Further, in the present embodiment, uppercase letters of the alphabet are converted into lowercase letters, and then the dictionary search in the normalization processing is performed.

FIG. 14 is a schematic diagram of the index when 1 is specified as the value of the normalization parameter at the time of registration. According to the normalization process shown in FIG. 8, "classified" and
"shopping" (Document A), "accounting" and "hired"
(Document C) and "classifying" (Document D) are representative notations "classify", "shop", "account",
Since "hire" and "classify" are returned, these are registered as index words. 1 is registered as the value of the search constraint flag in the normalized index word.

FIG. 15 is a schematic diagram of the index when 2 is specified as the value of the normalization parameter at the time of registration. According to the normalization process shown in FIG. 9, the heading notation is returned together with the representative notation, so these are registered as index terms. Specifically, in addition to the index in Fig. 14, "accou
nting "," classified "," classifying "," hired ","
"shopping" is registered.

FIG. 16 is a schematic diagram of the index when 3 is specified as the value of the normalization parameter at the time of registration. According to the normalization process shown in FIG. 11, since the endings are returned together with the representative notation, these are registered as index words. Specifically, in addition to the index in Fig. 14, "d", "in
"g" and "ping" are registered. 2 is registered as the value of the search constraint flag in the ending notation.

The number of different index words in each index is 10 in FIG. 14, 15 in FIG. 15, and 14 in FIG. In the scale of this embodiment, there is almost no difference in the index size between the case where the normalization parameter is 2 and the case where the normalization parameter is 3, but
The difference becomes more remarkable as the target data becomes larger. For example, if the target data contains 10,000 plural nouns and the normalization parameter is 2, it is necessary to register the same number of representative notations as that.
When the normalization parameter is 2, the index words registered as endings are limited to the four notations "s", "es", "ses", and "zes".

Next, a specific example of the document search process according to this embodiment will be described. FIG. 17 is an example of an inquiry text input through the input processing unit 10. "Account" in the first query and "acco" in the second query
unting "is" classified "in the third inquiry
However, "ping" is input in the fourth inquiry.

FIG. 18 shows an example of search conditions created by the search condition creating unit 13 for each inquiry text. The normalization parameter at the time of registration and the normalization mode at the time of search are specified by any of the following. That is the case. (1) Normalization parameter = 1. (2) Normalization parameter = 2, normalization mode. (3) Normalization parameter = 3, normalization mode.

For the first and second search conditions, FIG.
Index word "accoun" in each index from Fig. 16 to Fig. 16
When t "matches, documents B (ID = 2) and C (ID = 3) are searched. For the third search condition, the index word" classify "in each index shown in FIGS. And the documents A (ID = 1) and D (ID = 4) are searched for. With respect to the fourth search condition, “ping” matches only in FIG. Since it is registered, this is excluded and the number of searches is zero.

FIG. 19 shows that for each inquiry text,
It is an example of the search condition when the normalization parameter at the time of registration is 2 and the non-normalization mode is specified at the time of search. According to the processing of FIG. 10 (N in step S60), the same notation as the inquiry is used as the search condition.

The index word "account" in the index of FIG. 15 matches the first search condition, but the document ID
In the case of No. 3, since 1 is registered in the search constraint flag, this is excluded and only the document B (ID = 2) is searched. For the second search condition, the index word "accounting" in the index of FIG. 15 matches, and the document C (ID
= 3) is searched. For the third search condition, see FIG.
The index word "classified" in the index of
Document A (ID = 1) is searched. With respect to the fourth search condition, “ping” matches only in FIG. 16, but since 2 is registered in the search constraint flag, this is excluded and the number of searches is zero.

FIG. 20 shows that for each inquiry text,
It is an example of the search condition when the normalization parameter at the time of registration is 3 and the non-normalization mode is specified at the time of search. For the first inquiry, since there is no matching heading in the dictionary (N in step S81 in FIG. 12), the same notation as the inquiry is used as the search condition. For the second query, there is a matching heading in the dictionary (Y in step S81 in FIG. 12) and the corresponding ending field is described (Y in step S84 in FIG. 12), so both the representative notation and the ending Is used as the search condition. In the present embodiment, in such a case, the search condition is created using the operator #WORD (A, B). This operator takes a document in which arguments A and B appear at the same appearance position as a search result. Regarding the third inquiry, although there is a matching heading in the dictionary (Y in step S81 in FIG. 12), but the corresponding ending field is not described (N in step S84 in FIG. 12), the same notation as the inquiry is used. Used for search conditions.

Here, for the first search condition, FIG.
If the index word "account" in the index of the document matches, but the document ID is 3, the search constraint flag is 1
Has been registered, this is excluded and document B (ID =
Only 2) is searched. For the second search condition,
Since the index words "account" and "ing" in the index of 5 match and they appear at the same appearance position, the document C (ID
= 3) is searched. For the third search condition, see FIG.
The index word "classified" in the index of
Document A (ID = 1) is searched. With respect to the fourth search condition, “ping” matches only in FIG. 16, but since 2 is registered in the search constraint flag, this is excluded and the number of searches is zero.

As described above, when the normalization parameter at the time of registration is 2 or 3, whether or not the inflectional form and the representative word form are regarded as the same word to be searched can be specified at the time of search, so that a flexible search can be performed. It is possible. Comparing the two in terms of search efficiency, first, when the search condition is created, if the parameter is 2, as shown in FIG. 10, in the denormalization mode, processing such as dictionary search need not be performed at all. On the other hand, when the parameter is 3, as shown in FIG. 12, a dictionary search is required even in the denormalization mode. Further, also in the search process, when the parameter is 2, only the value of the search constraint flag is checked for the matched index word, while when the parameter is 3, the #WORD operator is evaluated. As a result, the search efficiency is generally low.

As described above, in the document search device 1 of the present embodiment, according to the normalization parameter at the time of registration, the conventional inflectional form and the typical word form are always (or not) regarded as the same word. It is possible to perform both flexible searches that can be specified at the time of search as to whether or not to regard them as the same word. That is,
The user can flexibly select at the time of search whether or not to search by considering the representative word form and the inflectional form as the same word while avoiding the efficiency reduction due to the search word expansion. Furthermore, regarding the latter,
It is possible to select either the method of giving priority to the index size or the method of giving priority to the search efficiency.

Further, in the present embodiment, the system of the stand-alone environment is shown as the document retrieval apparatus 1, but the present invention is not limited to this, and the client / server
The document search device 1 may be constructed as a system. In this case, the document search device 1 is provided on the server side,
Input of search conditions and the like from the input unit 5 of the client is transmitted to the input processing unit 10 of the document search apparatus 1 on the server side, and the search result is returned from the output processing unit 15 to the requesting client. The client is configured to display the result on the display unit 6 such as a display.

Further, in the document searching apparatus according to the first embodiment of the present invention, the text processing unit 11 does not consider the inflectional form extracted from the inquiry text and the representative word form in the document as the same word. When searching, the registration processing unit 12
Since the representative word form registered by does not match the search word, more flexible search can be performed.

Further, in the document searching apparatus according to the first embodiment of the present invention, the search processing unit 14 is configured so that the difference character string registered by the registration processing unit 12 does not match the search word by itself. Therefore, it is possible to avoid a match between a partial character string that does not stand alone as a word and a search word, and reduce search noise.

Further, since the document searching apparatus according to the first embodiment of the present invention enables the user to specify the method of registering the inflection type index by the registration processing unit 12 from the input unit 5 or the like, Depending on the needs of the user, it is possible to select whether to prioritize improvement in search efficiency or memory saving.

In the first embodiment, the case where the program for realizing the above-described document search method (including FIGS. 3 to 6 and 8 to 12) is stored in the memory 3 has been described. The invention is a computer-readable recording medium (CD-ROM, FD, magneto-optical disk (MO), mini-disk (MD), in which, in addition to the memory 3, a program for realizing the above-described document retrieval method is recorded. Even if a rewritable CD (including CD-RW) is attached to the document retrieval device 1, the recording medium is read by the FDD 7, the CD-ROM drive 8, etc. and the program is executed, the same effect can be obtained. Is. With this configuration, the program can be easily updated by moving and exchanging the recording medium.

Further, in the first embodiment, the case where the program for realizing the above-described document search method (including FIGS. 3 to 6 and 8 to 12) is stored in the memory 3 has been described. In addition to the invention,
The same effect can be obtained by downloading the program from the external device such as N to the memory 3 and executing it. By this method, the program can be easily updated via the network.

Here, the document storage means includes a document storage section 16 and the like, the inquiry input means includes an input section 5 and an input processing section 10, and the character string information extraction means includes a text. The index storage means includes a text processing section 11, a registration processing section 12, an index storage section 17, a data storage section 18, etc., and the search condition creation section includes a search condition creation section 13 Etc., the search processing means includes a search processing unit 14 and the like, and the designation means includes an input unit 5, a CPU 2, a memory 3 and the like.

[Second Embodiment] The document retrieval apparatus according to the second embodiment of the present invention has a hardware configuration substantially similar to that of the first embodiment. The same reference numerals are given and the description is omitted. Also,
The functional configuration of this embodiment is the same as the first configuration except for the list storage unit 19.
Since it is almost the same as that of the embodiment described above, the same components are given the same reference numerals and the description thereof will be omitted.

Here, FIG. 21 is a block diagram showing the functional structure of the essential parts of the document search device 1 according to the second embodiment of the present invention. In FIG. 21, the list storage unit 19 stores a non-search ending list (shown in FIG. 23).

Next, the document search method according to this embodiment will be described. Document registration processing in the registration processing unit 12, character string information extraction processing at the time of registration in the text processing unit 11, document search processing in the search condition creation unit 13, search processing unit 14, and output processing unit 15, and text Processing unit 11
Regarding the character information extraction processing at the time of searching in
Since it is substantially the same as the embodiment (shown in FIGS. 3 to 6), the description thereof will be omitted.

Here, the normalization process will be described. In addition, in the present embodiment, the following four values are set as the parameters of the normalization process at the time of index registration. 0: Normalization processing is not performed. 1: Perform normalization processing and register only the stem. 2: Normalize and register both stem and inflection. 3: Normalize and register the stem and ending.

In the present embodiment, further, at the time of search, whether the stem and the inflected form are regarded as the same word (normalized mode) or not regarded as the same word (non-normalized mode) is determined. Furthermore, it is assumed that whether to use the non-search list is specified. Here, the parameter and each mode are
The user may specify it. The parameters and each mode designated by the user are input, for example, by the input unit 5 and are stored in the hard disk 4 or the memory 3 in an updatable manner. Correspondence between the parameter value at the time of registration and the feasibility of each mode at the time of search is
It looks like this: 0: Only non-normalized mode is possible. 1: Only normalization mode is possible. 2: Both modes are possible. 3: Both modes are possible, non-search ending list can be used.

FIG. 22 is a description example of the ending normalization data stored in the data storage unit 18. In this embodiment, dictionary format data for word notation is used. The dictionary data is composed of three fields, the first field is a headline, and the inflection form to be normalized is described. The stem of the headline is described in the second field, and the ending is described in the third field.

In FIG. 22, the description of the ending is composed of a plurality of endings and is divided into a plurality of endings with a comma (,) as a delimiter. this is,
The invention according to claim 6 is applied. For example, "c
"lassification" is the continuation of the stem "class" and the endings "ify" and "tion", and "legalizing" is the stem "legal" and the endings "ize" and "in"
It is described as a series of g ". When it is described without being decomposed into multiple endings, for example," classification "is the stem" class ".
And the endings "ification" and "legalizing" are described as the stem "legal" and the ending "izing". This is an application of the invention according to claim 4.

FIG. 23 is a description example of the non-search ending list stored in the list storage section 19. Here, "ed", "ing",
The four endings of "s" and "tion" are listed.

FIG. 24 is a flow chart showing an outline of the normalization process at the time of registration and search when the value of the normalization parameter at the time of index registration is 1. The text processing unit 11 performs a dictionary search using the index word or the search word as a key (step S90), and if there is no matching heading (N in step S91), the process ends. If there is a matching headline (Y in step S91), the corresponding stem is returned (step S92), and the process ends.

FIG. 25 is a flow chart showing the outline of the normalization process at the time of registration when the value of the normalization parameter at the time of index registration is 2. The text processing unit 11 performs a dictionary search using the index word as a key (step S10).
0) If there is no matching headline (N in step S101), the process ends. If there is a matching headline (Y in step S101), both the headline and the stem are returned (step S102), and the process ends.

FIG. 26 is a flow chart showing an outline of the normalization process at the time of search when the value of the normalization parameter at the time of index registration is 2. If the parameter value is 2 or more, it is possible to specify the normalization mode or the non-normalization mode. Therefore, if the text processing unit 11 is not the normalization mode (N in step S200), the process ends. If it is in the normalization mode (Y in step S200), a dictionary search is performed using the search word as a key (step S201). If there is no matching headline (N in step S202), the process ends. If there is a matching headline (Y in step S202), the corresponding stem is returned (step S203), and the process ends.

FIG. 27 is a flow chart showing an outline of the normalization process at the time of registration when the value of the normalization parameter at the time of index registration is 3. The text processing unit 11 performs a dictionary search using the index word as a key (step S30
0) If there is no matching headline (N in step S301), the process ends. If there is a matching headline (Y in step S301), both the headline and the stem are returned (step S302), and the process ends.

FIG. 28 is a flow chart showing an outline of the normalization process at the time of search when the value of the normalization parameter at the time of index registration is 3. The text processing unit 11 first performs a dictionary search using the search word as a key (step S
400), if there is no matching headline (N in step S401), the process ends. If there is a matching headline (Y in step S401), the word stem is returned (step S402), it is determined whether or not the mode is the normalization mode, and if it is in the normalization mode (Y in step S403), the process ends. If it is not the normalization mode (N in step S403), it is confirmed whether or not the ending field corresponding to the headline is described. If it is not described (N in step S404), the process ends. If it is described (Y in step S404), it is confirmed whether or not the non-search ending list is used. If not used (N in step S405), all endings are returned (step S407). If used, (Y in step S405),
The endings described in the list are excluded (step S40
6) The remaining endings are returned (step S407), and the process ends.

Next, a specific example of the document registration process according to this embodiment will be described. FIG. 29 is a schematic diagram of a document to be registered, and in this embodiment, there are six documents A ′ to F ′.
The document A ′ is English document data including a character string “Classified Ad from Ed” in the document, and is given an identifier (document ID: 1 ′) that uniquely indicates the document A ′.
The document B ′ is English document data containing a character string “class reunion” in the document, and is given an identifier (document ID: 2 ′) that uniquely indicates the document B ′. Document C '
Is English document data containing a character string "target classification" in the document, and is given an identifier (document ID: 3 ') that uniquely indicates the document C'. Document D '
Is English document data containing a character string "opposition to legalizing drugs" in the document, and is given an identifier (document ID: 4 ') uniquely indicating the document D'.
Document E'includes "opposed legalization of drug" in the document.
The document data is an English document data including a character string "s", and an identifier (document ID: 5 ') that uniquely identifies the document E'is added. The document F'is included in the document "legality of random d.
This is English document data including a character string "rug testing", and an identifier (document ID: 6 ') uniquely indicating the document F'.
Is given.

Next, the value of the normalization parameter at the time of registration is 1
The indexes in the cases of 1 to 3 will be described. If the parameter value is 0, the word form in the document is registered as it is, and therefore the description is omitted. The index is composed of an index word notation, an ID given to a document in which the index word appears, an appearance position in each document, and a search constraint flag indicating a constraint at the time of search.

The search constraint flag is 0 when there is no constraint, 1 when the index word is the stem obtained by the normalization processing, and 1 when the index word is the ending obtained by the normalization processing. 2 are registered respectively. When the search constraint flag is 1, it is not a search target by itself in the non-normalization mode, and when the search constraint flag is 2, it is a search target by itself in both the normalization mode and the denormalization mode. It should not be. Further, in the present embodiment, uppercase letters of the alphabet are converted into lowercase letters, and then the dictionary search in the normalization processing is performed.

FIG. 30 is a schematic diagram of the index when 1 is specified as the value of the normalization parameter. According to the normalization process shown in FIG. 24, "classified" (document A '), "classification" (document C'), "opositio"
n "," legalizing ", and" drugs "(Document D ')," oppos
"ed", "legalization", and "drugs" (Document E '), "le"
For "gality" and "testing" (Document F '), the stem "clas"
Since s "," drug "," oppose ", etc. are returned, these are registered as index words.In the normalized index word, 1 is registered as the value of the search constraint flag.

FIG. 31 is a schematic diagram of the index when 2 is specified as the value of the normalization parameter. According to the normalization processing shown in FIG. 25, the stem and the headline are returned, so these are registered as index words. Specifically, in addition to the index when the parameter value = 1, "classif
ication "," classified "," classifying "," drugs ",
And "legality" etc. are registered.

FIG. 32 is a schematic diagram of an index when 3 is specified as the value of the normalization parameter. According to the normalization process shown in FIG. 28, the stem and ending are returned,
These are registered as index terms. Specifically, in addition to the index when the parameter value = 1, "ed", "if
“Y”, “ing”, “ize”, “s”, and “tion” are registered, and 2 is registered as the value of the search constraint flag in the word suffix.

Here, the number of different index words in each index is 12 when the parameter value = 1, 23 when the parameter value = 2, and 18 when the parameter value = 3. The number of characters in different index terms is 48 when the parameter value = 1 and 149 when the parameter value = 2.
It is 65 when the parameter value = 3. Parameter value =
Although not as great as in the case of 1, when the parameter value = 3,
Compared to the case where the parameter value = 2, a considerably smaller index size can be realized.

Next, a specific example of the document search process according to this embodiment will be described. 1. Searching for documents related to drug legalization with the inquiry text "drug legalization" 1a. If the normalization parameter at the time of registration = 1, only the normalization mode can be used as the search mode.
The inflectional form "legalization" of the inquiry text is always normalized to "legal", and search conditions such as #AND (drug, legal) are created, and documents D '(ID = 4') and E '(ID =
5 ') and F' (ID = 6 ') are searched. Of these documents, document F'is a document relating to "lawfulness of randomized drug tests", which may be unrelated to "legalization of drugs", but document F'cannot be searched. Can not. 1b. When the normalization parameter at registration = 2, 1 when the normalization mode is selected as the search mode
a. Is the same as. If you choose denormalized mode,
A search condition such as #AND (drug, legalization) is created, and only the document E ′ (ID = 5 ′) is searched. Document D '(ID = 4'), which appears to be related to "legalization of drug", is not retrieved. 1c. Normalization parameter at registration = 3 When normalization mode is selected as the search mode, 1
a. Is the same as. If you select the denormalization mode and do not use the non-search ending list, #AND (drug, # WORD (le
gal, ize, tion)) is created and 1b.
Similar to the above, only the document E '(ID = 5') is searched, and the document D '
(ID = 4 ') is not searched. Here, operator #WORD
(A, B, ...) The search results are documents in which the arguments A, B, ... (two or more search words) appear at the same appearance position. In addition, when the non-normalization mode is selected and the non-search ending list is used, the ending "tion" is excluded, so a search condition such as #AND (drug, # WORD (legal, ize)) Created, document D '(ID = 4'), document E '(ID = 5')
Is searched, the document F ′ (ID = 6 ′) is not searched, and the search result as specified by the user is obtained.

2. In the inquiry text "classfied",
When searching for documents related to three-line advertising 2a. When the normalization parameter at the time of registration = 1, since only the normalization mode is possible as the search mode, "classified" is always normalized to "class", and the documents A '(ID = 1') and B '(ID = 2 ') and C' (ID = 3 ') are searched. Of these, only A'is the document related to the three-line advertisement, but documents B'and C'cannot be searched. 2b. When the normalization parameter at registration = 2, 2 when the normalization mode is selected as the search mode
a. Is the same as. If you choose denormalized mode,
Only document A '(ID = 1') is retrieved. 2c. When the normalization parameter at the time of registration = 3, 2 when the normalization mode is selected as the search mode
a. Is the same as. If you select denormalization mode and do not use non-search ending list, #WORD (class, ify, e
A search condition as shown in d) is created, and 2b. Similar to the above, only the document A ′ (ID = 1 ′) is searched.

3. Searching for documents related to class meetings using the inquiry text "class" 3a. When the normalization parameter at the time of registration = 1, the document related to the class meeting is only the document B ', but the documents A' (ID = 1 '), B' (ID = 2 '), C' (ID = 3 ') are always To be searched. 3b. When the normalization parameter at the time of registration = 2, 3 when the normalization mode is selected as the search mode
a. Is the same as. If you choose denormalized mode,
Since the processing is performed so that only the document whose search constraint flag corresponding to the index word "class" is 0 is searched, the document B '
Only (ID = 2 ') is searched. 3c. When the normalized parameter at the time of registration = 3 In this case, 3b. Is the same as.

4. Searching for documents related to the personal name "Ed" in the inquiry text "Ed" 4a. In the case of normalization parameter = 1 at the time of registration In any search mode, only the document A ′ (ID = 1 ′) is searched. 4b. When the normalization parameter at the time of registration = 2 In this case, 4a. Is the same as. 4c. When the normalization parameter at the time of registration = 3, there are documents E ′ (ID = 5 ′) in addition to document A ′ (ID = 1 ′) as documents that include “ed” as an index word. Since the search constraint flag for the index word "ed" is 2, no search is performed in any search mode, and the document A '(ID =
Only 1 ') are searched. As described above, when the normalization parameter at the time of registration is 2 or 3, whether or not to search by considering the inflectional stems as the same word can be specified at the time of search, so that flexible search is possible. Comparing the two in terms of search efficiency, first, when the parameter is 2 when the search condition is created, as shown in FIG. 26, in the denormalization mode, it is not necessary to perform processing such as dictionary search at all. On the other hand, when the parameter is 3, as shown in FIG. 28, a dictionary search is required even in the denormalization mode. Furthermore,
Also in the search process, if the parameter is 2, the value of the search constraint flag is only checked for the matched index word, whereas if the parameter is 3, the #WORD operator is evaluated. , Overall, search efficiency is low.

As described above, in the document search device 1 according to the second embodiment of the present invention, the conventional inflectional form and the typical word form are always regarded as the same word according to the normalization parameter at the time of registration ( It is possible to perform both a search (or not to consider) and a flexible search in which whether or not to consider the same word can be specified at the time of search. Further, with regard to the latter, it is possible to select either a method giving priority to the index size or a method giving priority to the search efficiency. That is, while avoiding a decrease in efficiency due to search word expansion, the user can flexibly search at the time of search by using the input unit 5 or the like as to whether or not to search by considering the stem and the inflectional form or the inflectional forms as the same word. You can choose.

In the present embodiment, the case where the document retrieval apparatus 1 is applied to a stand-alone environment is shown, but the present invention is not limited to this, and the client / server
The document search device 1 may be constructed as a system. In that case, the document search device 1 is provided on the server side, and inputs such as search conditions from the input unit 5 of the client are transmitted to the input processing unit 10 of the document search device 1 in the server, and the search result is output from the output processing unit 16. Reply to the requesting client. The client is configured to display the result on the display unit 6 including a display or the like.

Further, the document retrieval apparatus 1 according to the second embodiment of the present invention, when considering the inflection form extracted from the inquiry text and the stem character string in the document data as the same word, retrieves them. Uses only the stem character string obtained by decomposing the extracted character string by the search processing unit 14 as a search condition, so that the stem and the inflected form, or the inflected forms are regarded as the same word for searching. As a result, omission of search can be avoided.

Further, the document retrieval apparatus 1 according to the second embodiment of the present invention retrieves the inflection form extracted from the inquiry text and the stem character string in the document data without considering them as the same word. In this case, the search processing unit 14 uses the stem character string and the ending character string obtained by decomposing the extracted character string as the search condition, so that the stem and the ending variation form and the ending variation form are the same. Search noise can be avoided by searching without considering it as a word.

Further, in the document search device 1 according to the second embodiment of the present invention, the search processing unit 14 does not always use a preset ending character string as a search condition, so that the search noise A highly accurate search can be performed by excluding from the search conditions only the ending character strings that are unlikely to occur.

Further, in the document searching apparatus 1 according to the second embodiment of the present invention, the search processing unit 14 prevents the suffix character strings registered by the registration processing unit 12 from being searched independently. By avoiding that the end word string that does not hold as a word and the search word match,
Search noise can be reduced.

Further, in the document searching apparatus 1 according to the second embodiment of the present invention, the user inputs the registration processing unit 12 with the registration method of the inflection type index storage unit 17 extracted from the document data. Since the designation can be made from the section 5 or the like, it is possible to select which of the improvement of the search efficiency and the memory saving is prioritized according to the needs of the user.

Further, in the document searching apparatus 1 according to the second embodiment of the present invention, the user can specify, through the input unit 5 or the like, the method for searching the inflectional form extracted from the inquiry text by the registration processing unit 12. Since this is done, it is possible to perform a flexible search according to the designation of the user.

Further, in the document search device 1 according to the second embodiment of the present invention, the search processing unit 14 allows the user to input whether the preset suffix character string is always used as a search condition. Since it can be specified from the etc., more flexible search can be performed according to the user's specification.

In the second embodiment, the case where the program for realizing the above-described document search method (including FIGS. 24 to 28) is stored in the memory 3 has been described, but the present invention is not limited to the memory 3. A computer-readable recording medium (CD-ROM, FD, magneto-optical disk (MO), in which a program for realizing the document retrieval method described above is recorded.
Even if a mini disk (MD), rewritable CD (including CD-RW) is attached to the document retrieval device 1 and this recording medium is read by the FDD 7, the CD-ROM drive 8 etc. and the program is executed, the same applies. The effect of is obtained. With this configuration, the program can be easily updated by moving and exchanging the recording medium.

Further, in the second embodiment, the case where the program for realizing the above-described document search method (including FIGS. 24 to 28) is stored in the memory 3 has been described. However, the present invention is not limited to this. Even if the communication unit 20 downloads the program from the external device on the network such as LAN to the memory 3 and executes the program, the same effect can be obtained. By this method, the program can be easily updated via the network.

Here, the document storage means includes the document storage section 16 and the like, the inquiry input means includes the input section 5, the input processing section 10 and the like, and the character string information extraction means includes the text. The index storage means includes a processing section 11, etc., and the index storage means includes a text processing section 11, a registration processing section 12, an index storage section 17, a data storage section 18, a non-search word ending list 19, etc. The search processing means includes a search processing unit 14 and the like.

[0093]

As described above, according to the present invention, according to the normalization parameter at the time of registration, the conventional inflectional form and the representative form are always regarded as the same word, or the search is made not to be regarded as the same word. By selecting and executing one of a search that can be designated at the time of search as to whether or not to regard it as the same word, regardless of whether or not the search word that is the inflected form is normalized to the representative word form, It is possible to provide a document search device having an excellent effect that a document in which a search word appears can be searched.

[Brief description of drawings]

FIG. 1 is a block diagram showing a hardware configuration of essential parts of a document search device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a functional configuration of main parts of the document search device according to the first embodiment of the present invention.

FIG. 3 is a flowchart showing an outline of document registration processing of a registration processing unit according to the first and second embodiments of the present invention.

FIG. 4 is a flowchart showing an outline of character information extraction processing (during registration) of the text processing unit according to the first and second embodiments of the present invention.

FIG. 5 is a flowchart showing an outline of document search processing of a search condition creation unit, a search processing unit, and an output processing unit according to the first and second embodiments of the present invention.

FIG. 6 is a flowchart showing an outline of character information extraction processing (at the time of search) of the text processing unit according to the first and second embodiments of the present invention.

FIG. 7 is a diagram illustrating a description of ending normalization data according to the first embodiment of the present invention.

FIG. 8 is a flowchart showing an outline of a normalization process at the time of registration and search (when the value of a normalization parameter at index registration is 1) according to the first embodiment of the present invention.

FIG. 9 is a flowchart showing an outline of a normalization process at the time of registration (when the value of a normalization parameter at the time of index registration is 2) according to the first embodiment of the present invention.

FIG. 10 is a flowchart showing an outline of normalization processing at the time of search (when the value of the normalization parameter at the time of index registration is 2) according to the first embodiment of the present invention.

FIG. 11 is a flowchart showing an outline of a normalization process at the time of registration (when the value of a normalization parameter at the time of index registration is 3) according to the first embodiment of the present invention.

FIG. 12 is a flowchart showing an outline of a normalization process at the time of search (when the value of the normalization parameter at the time of index registration is 3) according to the first embodiment of the present invention.

FIG. 13 is a schematic diagram of a registration target document according to the first embodiment of the present invention.

FIG. 14 is an index according to the first embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 1).
FIG.

FIG. 15 is an index according to the first embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 2).
FIG.

FIG. 16 is an index according to the first embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 3).
FIG.

FIG. 17 is a diagram illustrating an inquiry text input through the input processing unit according to the first embodiment of the present invention.

FIG. 18 is a diagram illustrating search conditions (including classify) created by a search condition creation unit according to the first embodiment of the present invention.

FIG. 19 is a diagram illustrating search conditions created by a search condition creation unit according to the first embodiment of the present invention.

FIG. 20 is a diagram illustrating search conditions (including #WORD) created by a search condition creation unit according to the first embodiment of the present invention.

FIG. 21 is a block diagram showing a functional configuration of main parts of a document search device according to a second embodiment of the present invention.

[Fig. 22] Fig. 22 is a diagram for describing a description of word end normalized data according to the second embodiment of the present invention.

FIG. 23 is a diagram illustrating a description of a non-search ending list according to the second embodiment of the present invention.

FIG. 24 is a flowchart showing an outline of normalization processing at the time of registration and search (when the value of a normalization parameter at index registration is 1) according to the second embodiment of the present invention.

FIG. 25 is a flowchart showing an outline of a normalization process at the time of registration (when the value of a normalization parameter at the time of index registration is 2) according to the second embodiment of the present invention.

FIG. 26 is a flowchart showing an outline of normalization processing at the time of search (when the value of the normalization parameter at the time of index registration is 2) according to the second embodiment of the present invention.

FIG. 27 is a flowchart showing an outline of a normalization process at the time of registration (when the value of a normalization parameter at the time of index registration is 3) according to the second embodiment of the present invention.

FIG. 28 is a flowchart showing an outline of a normalization process at the time of search (when the value of the normalization parameter at the time of index registration is 3) according to the second embodiment of the present invention.

FIG. 29 is a schematic diagram of a registration target document according to the second embodiment of the present invention.

FIG. 30 is an index according to the second embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 1).
FIG.

FIG. 31 is an index according to the second embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 2).
FIG.

FIG. 32 is an index according to the second embodiment of the present invention (when the value of the normalization parameter at the time of index registration is 3).
FIG.

[Explanation of symbols]

1 Document retrieval device 2 CPU 3 memory 4 hard disk 5 Input section 6 Display 7 FDD 8 CD-ROM drive 9 buses 10 Input processing unit 11 Text processing section 12 Registration Processing Department 13 Search condition creation section 14 Search processing section 15 Output processing unit 16 Document storage 17 Index storage 18 Data storage 19 List storage 20 Communications department

Claims (6)

[Claims] 1. A document data, which is divided into words and described in a language having a ending change, and which is digitized,
Document storage means for assigning an identifier to each document for storage, inquiry input means for inputting inquiry text for searching a desired document from the document data stored by the document storage means, and storage by the document storage means Character string information for extracting the document data or the inquiry text input by the inquiry input unit, and character string information including the character string delimited by a predetermined delimiter as a boundary and the appearance position of the character string Extraction means, in the document data stored by the document storage means, information including the identifier of the document in which the character string extracted by the character string information extraction means appears and the appearance position of the character string are associated with the character string. And store it in the index, and when the character string is inflected, the representative word form for the character string is Index storage means for storing in the index in association with the character string, and search in a predetermined format from the inquiry text input by the inquiry input means based on the character string extracted by the character string information extraction means Search condition creating means for creating a condition, and referring to the index stored by the index storage means,
And a search processing unit that searches for a document satisfying the search condition created by the search condition creating unit. 2. The index storage means, when the character string extracted by the character string information extracting means is an inflection, and the inflection completely includes a representative word form for the character string. 3. The character string different from the representative word form and the representative word form in the inflectional form is stored in the index in association with the same appearance position information.
Document retrieval device described in. 3. When the character string extracted from the inquiry text by the character string information extracting means is inflected, the search processing means makes the inflected and the representative word in the document the same. Designation means for designating whether or not to search by considering it as a word, or whether or not to search by considering the representative word form of the character string extracted from the inquiry text and the inflection form in the document as the same word The document search device according to claim 1, wherein the document search device is provided. 4. A document data which is divided into words and described in a language having a ending change and digitized,
Document storage means for giving an identifier to each document and storing the data, inquiry input means for inputting inquiry text for searching a desired document from the document data stored by the document storage means, and storage by the document storage means Character string information for extracting character string information delimited by a predetermined delimiter as a boundary, and character string information including the appearance position of this character string, for the document data that has been input or the inquiry text that has been input by the inquiry input means. Extraction means, and the document data stored by the document storage means, the information including the identifier of the document in which the character string extracted by the character string information extraction means appears and the appearance position of the character string, corresponding to the character string The character string that is attached and stored in the index, and that is extracted from the document data by the character string information extracting means is In the case of a tail change form, an index storage unit that stores a stem character string obtained by decomposing the character string into a stem and a word tail portion in an index in association with the same appearance position information as the character string, Search condition creating means for creating a search condition of a predetermined format from the inquiry text input by the inquiry input means, based on the character string extracted by the character string information extracting means, and the index storage means Referring to the index stored by
And a search processing unit that searches for a document satisfying the search condition created by the search condition creating unit. 5. The index storage means obtains the character string by decomposing the character string into a stem and a word ending portion when the character string extracted from the document data by the character string information extracting means is in the inflection form. 5. The document search device according to claim 4, wherein the generated stem character string and ending character string are stored in the index in association with the same appearance position information. 6. The index storage means further decomposes the ending character string when the ending character string obtained by decomposing the character string into a stem and a ending portion is composed of a sequence of a plurality of endings. The document search device according to claim 5, wherein the plurality of ending character strings and the stem character strings obtained by the above are stored in the index in association with the same appearance position information.