JP4014417B2 - Full-text search device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a full-text search device, and more particularly to a full-text search device that searches a document including a character string designated from a plurality of document data. The present invention can be applied to a system that manages a large amount of document data such as a document management system, an electronic library system, and a patent publication search system.
[0002]
[Prior art]
2. Description of the Related Art In recent years, documents that have been digitized due to the development of information communication technology and information related to the documents have been distributed in large quantities via the Internet or the like. A document search apparatus that searches a desired document with high accuracy and at a high speed during the distribution of the digitized document and information has been proposed.
[0003]
In such a document search apparatus, a keyword search method or a full-text search method is used. A full-text search device using a full-text search method is a device that performs matching between an arbitrary search character string and all documents to be searched, and extracts documents including the search character string without omission. In this way, a great deal of human power is not required such as assigning keywords in advance to all documents to be searched. Various types of full-text search devices have been proposed. One type of full-text search device is a device that employs a transposed (index) file method. In the transposed file method, as an auxiliary file for searching, a document in which characters / words / n-grams (n-character concatenation) appear, or a transposed file that records the appearance position in those documents is constructed in advance, and the whole text At the time of retrieval, the retrieval is performed using only the transposed file, and a very high-speed retrieval can be performed, which is effective for a system that requires a high-speed retrieval of a large number of documents.
[0004]
For details on the full-text search method in general and the transposed file method, refer to the conventional technology disclosed in Japanese Patent Laid-Open No. 11-073429, and the activity report of the full-text search system council in 1998 (http://www.ftsanet.com/dbtokyo99/Db99. htm), etc., and since it is publicly known, its description is omitted.
[0005]
However, with the inverted file method, it is usually necessary to construct an inverted file that is several times the original data, and the full-text index of the inverted file method requires more time for registration / deletion processing as the amount of registered document data increases. As a result, the response time of the registration / deletion process as viewed from the user side becomes longer for the full-text search apparatus. During the registration / deletion process, the search process must wait.
[0006]
Japanese Patent Application Laid-Open No. 7-146880 discloses a document search apparatus and method for registering a new document in a sub-index smaller than the main index and shortening the registration time. However, in the invention described in the publication, search processing cannot be performed during registration of a new document, although the registration time is short.
[0007]
[Problems to be solved by the invention]
The present invention has been made in view of the above circumstances, shortening the response time of registration and deletion processing as seen from the user side, and further performing search processing before registration processing and deletion processing are completed. It is an object of the present invention to provide a possible full-text search device.
[0008]
[Means for Solving the Problems]
According to the first aspect of the present invention, in a full-text search apparatus for searching for a document including a specified character string from a plurality of document data, a document data storage unit for storing registered document data, a full-text index storage unit for search, , An input means for inputting data from the user, an output means for outputting search results, a registration processing means for performing registration processing for document data, a deletion processing means for performing deletion processing for document data, and a search for performing search processing A full-text index storage unit for registration and a full-text index storage unit for deletion separately from the full-text index storage unit for search, and further, the full-text index storage unit for registration And data from the full-text index storage unit for deletion to the full-text index storage unit for search Merge to merge Means and a lock processing means for performing a lock process. When the data is merged from the full-text index storage unit for registration and the full-text index storage unit for deletion into the full-text index storage unit for search, the merge means transposes tokens that are constituent elements of the full-text index. Processing is performed for each list, and the lock processing means locks the tokens of the transposed list. It is characterized by that.
[0010]
Claim 2 The invention of claim 1 In the invention of merge When the number of document data registered in the registration full-text index storage unit reaches a predetermined number, the data is stored in the search full-text index storage unit. Merge It is characterized by performing processing.
[0011]
Claim 3 The invention of claim 1 In the invention of merge The means stores data in the full-text index storage unit for search when the capacity of the full-text index storage unit for registration reaches a predetermined capacity. Merge It is characterized by performing processing.
[0012]
Claim 4 The invention of claim 1 to claim 1 3 In any one of the inventions, merge When the number of document data registered in the deletion full-text index storage unit reaches a predetermined number, the data is stored in the search full-text index storage unit. Merge It is characterized by performing processing.
[0013]
Claim 5 The invention of claim 1 to claim 1 3 In any one of the inventions, merge The means stores data in the full-text index storage unit for search when the capacity of the full-text index storage unit for deletion reaches a predetermined capacity. Merge It is characterized by performing processing.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
In order to eliminate the length of response time of registration / deletion processing as seen from the user side in the transposed file method according to the prior art, the present applicant registers a small full-text index in Japanese Patent Application No. 2001-223604. Prepared separately for registration and deletion to prevent deterioration of response time for registration and deletion. For search processing, add the search result of small full-text index for registration to the search result of large-scale full-text index. We proposed a full-text search device that uses search results to be returned to users, except for small-scale full-text index search results. In this method, the technique described in Japanese Patent Application No. 2001-78026 by the present applicant is applied to a full-text search apparatus to prevent deterioration in response times of registration and deletion. In the invention described in the specification of Japanese Patent Application No. 2001-223604, the time required for data transfer is shortened by transferring an inverted list that is a constituent element of a full-text index. In the data transfer means from the full-text index to the large-scale full-text index, the time required for data transfer is shortened by using the transposed file type full-text index instead of using the original document data.
[0015]
In the above-mentioned Japanese Patent Application No. 2001-78026, a database management system, a program, and a program capable of further improving the update performance during system operation while maintaining the performance capable of responding to an advanced search request at high speed, and A recording medium is described, and a registration / deletion throughput is increased by preparing a data holding unit for registration / deletion separately from a data holding unit for search. However, according to the technique described in the above-mentioned Japanese Patent Application No. 2001-78026, registration is made in a small-scale index by means of data transfer from a small full-text index for registration and deletion to a large full-text index for search. The original document data is acquired from the identifier of the existing document data, and is registered and deleted in a large-scale index. As described above, registration / deletion processing to a large-scale full-text index takes time, so the data transfer processing time becomes longer, and generally search processing cannot be performed during registration / deletion processing to the full-text index. There is a problem that the response time of the search processing as seen from the user is deteriorated.
[0016]
In the invention described in the above-mentioned Japanese Patent Application No. 2001-223604, the search process cannot be performed until the transfer process of all the transposed lists is completed. That is, search processing cannot be performed unless data transfer from a small full-text index for registration and deletion to a large full-text index for search is completed. In the present invention, a lock process is added to a token so that a search process can be performed without waiting for the transfer process to end. In other words, in the present invention, in the data transfer means from the small full-text index to the large full-text index, the token of the transposed list to be transferred is locked so that the search can be performed even during the transposed list transfer.
[0017]
FIG. 1 is a block diagram for explaining the function of a full-text search device according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of a hardware configuration when the full-text search device in FIG. FIG. 3 is a diagram illustrating a hardware configuration example when the full-text search apparatus in FIG. 1 is configured by a server / client.
The full-text search device according to the present invention is a device for searching for a document including a specified character string from a plurality of document data (a plurality of digitized documents). Referring to FIG. 1, in the present embodiment, in the input means (input processing means) 1, data such as text data for registration processing, document identifiers for deletion processing, search conditions for search processing, and the like are input from the user. , Processing to pass to the registration processing means 3, the deletion processing means 4 and the search processing means 5, respectively. The registration processing means 3 performs registration processing relating to document data. Registration processing in the registration processing means 3 is performed for the document data storage unit 7 and the registration full-text index storage unit (often referred to as a registration small-scale full-text index storage unit) 9. The deletion processing unit 4 performs deletion processing related to document data. The deletion processing in the deletion processing unit 4 reads out the document data stored in the document data storage unit 7 on the basis of the document identifier input by the input unit 1 and uses the text dividing unit 6 to store a small full-text index for registration. If the index is registered in the section 9, the index is deleted, and if it is not a registered index, the index is recorded in the deletion full-text index storage section (often referred to as a small-scale full-text index storage section for deletion) 10. To do.
[0018]
In the text dividing means 6, the division processing from document data to partial character strings in registration processing and the conversion from document data to partial character strings required in each of the registration processing means 3, deletion processing means 4 and search processing means 5 are performed. The dividing process from the search condition (searched character string) to the partial character string is performed. The search processing in the search processing means 5 includes a search full-text index storage unit (often referred to as a large-scale search full-text index storage unit) 8, a small-scale full-text index storage unit 9 for registration, and a small-scale full-text index storage unit for deletion. 10, the result obtained by subtracting the search result in the storage unit 10 from the search results in the storage units 8 and 9 is obtained, and the result is output by the output unit 2 as the search result. The merging means 11 performs data transfer among the search large-scale full-text index storage unit 8, the registration small-scale full-text index storage unit 9, and the deletion small-scale full-text index storage unit 10. As a feature of the present invention, the lock processing means 12 performs a lock process for locking each process in each of the means 3, 4, 5 and 11 in order to prevent other processes. It can be said that the lock processing means 12 is a management means for managing lock processing.
[0019]
In the stand-alone hardware configuration shown in FIG. 2, the input means 1 in FIG. 1 is realized by the input device 21, and the output means 2 is realized by the display device 22. Various processing means 3 to 6, 11 and 12 are realized in a main control device (CPU, memory, etc.) 24, and various storage units 7 to 10 are realized in a storage device 25. The input / output control device 23 controls the input device 21 and the display device 22 in accordance with a control signal from the main control device 24.
[0020]
In the hardware configuration of the server / client shown in FIG. 3, the input means 1 in FIG. 1 is realized by the input device 31 of the client 30, and the output means 2 is realized by the display device 32 of the client 30. The various processing units 3 to 6, 11, and 12 are realized in the main control devices (CPU, memory, etc.) 34 and 52 of the client 30 and the server 50, and the various storage units 7 to 10 are realized in the storage device 53 of the server 50. . In addition, the network control devices 35 and 51 of the client 30 and the server 50 control data transmission and the like between the client 30 and the server 50 via the network 40. Further, the input / output control device 33 of the client 30 controls the input device 21 and the display device 22 according to the control signal of the main control device 34.
[0021]
Below, an example of operation | movement of the full-text search apparatus comprised as mentioned above is demonstrated in detail.
FIG. 4 is a flowchart for explaining the registration process in the full-text search apparatus according to the embodiment of the present invention.
To execute the registration process, the user first creates document data and registers (inputs) the document data from the input means 1 (step S1). Document data is stored in the document data storage unit 7 in the registration processing means 3 (step S3), and at the same time, an identifier (document identifier) indicating the document data is determined (step S2). Further, the registration processing means 3 uses the text dividing means 6 to obtain a partial character string (token) and appearance position information (transposition list) of the token from the document data (step S4). Next, an X lock is applied to the registration small transposed index storage unit (registration small full-text index storage unit) 9 by the lock processing means 12 (step S5). The X lock will be described later. Until the token is ended (during NO in step S6), the document identifier and the appearance position information of each token are recorded in the registration small full-text index storage unit 9 (step S7). That is, the transposed list is inserted into the registration small full-text index storage unit 9. If YES in step S6, the X-lock of the registration small full-text index storage unit 9 is released (step S8), and the process ends. The dividing method used by the text dividing unit 6 may be a method using N character sets as tokens, or a method using morpheme analysis and words as tokens. In the following example, the description will be made only for the text dividing means using the method of using the N character set as a token, but it can be similarly applied to the method of using the word subjected to morphological analysis as a token.
[0022]
FIG. 5 is a diagram for explaining processing in the full-text search apparatus of FIG. 1 and is a diagram illustrating an example of a full-text index. The full text index of the inverted file method will be described in detail using the example of FIG.
Registered document data is document 1 and document 2, and their contents (here, contents obtained by dividing by text dividing means 6) are respectively represented by reference numerals 61 and 62 in FIG. Here, the number on the left of each document represents the number of characters from the beginning of the character string. That is, in Document 1, “full text search” appears at the 11th character from the top, “method” appears at the 20th and 60th characters, and “full text search method” appears at the 31st character. In document 2, “search method” means the first character from the top, “method” appears in the 24th character, and “full text” appears in the 30th and 42nd characters.
[0023]
When using 2 character sets as partial character strings, all partial character strings in the document are extracted, and their appearance positions (number of characters from the beginning) are collectively recorded in the index for each partial character string. To do. For example, since “full text” appears at positions 11 and 31 and “sentence check” appears at positions 12 and 32 from document 1, it is recorded in the index. In the index, not only the appearance position in the document but also a document identifier for identifying in which document and the number of appearances are added and recorded, so the format is as shown by reference numeral 63 in FIG. For example, the transposed lists {1, 2, (11, 31)} and {2, 2, (30, 42)} for “full text” each appear twice in document 1 and their positions are 11, 31 , And it appears twice in document 2 and its position is 30,42.
[0024]
6 and 7 are flowcharts for explaining the deletion process in the full-text search apparatus according to the embodiment of the present invention.
In order to execute the deletion process, the user first inputs a document deletion request by inputting the document identifier of the document to be deleted from the input means 1 (step S11). Next, the lock processing means 12 applies X lock to the document (step S12), and the deletion processing means 4 reads out (takes out) the document data corresponding to the document identifier from the document data storage unit 7 (step S13). Further, the deletion processing unit 4 obtains a partial character string (token) and appearance information of the token from the document data using the text dividing unit 6 (step S14). Next, it is determined whether there is a document to be deleted in the registration small full-text index storage unit 9 (step S15). If the document identifier is not registered in the registration small full-text index (search large-scale full-text index) (No in step S15), an X lock is applied to the deletion small full-text index storage unit 10 (step S16), and the transposed list is stored until the token ends (YES in step S17). Inserted into the deletion small full-text index storage unit 10, that is, the document identifier and the appearance position information of each token are recorded in the deletion small full-text index storage unit 10 (step S18). The X lock is released (step S19). On the other hand, if the document identifier is a document identifier registered in the registration small full-text index (YES in step S15), the registration small full-text index storage unit 9 is X-locked (step S22) and the token is registered. Until the process ends (YES in step S23), the transposed list is deleted from the registration small full-text index storage unit 9, that is, the appearance position information of each token is deleted from the registration small full-text index storage unit 9 (step S24). The X-lock of the registration small full-text index storage unit 9 is released (step S25). Subsequent to steps S19 and S25, the deletion processing unit 4 deletes the document data corresponding to the document identifier from the document data storage unit 7 (step S20), releases the X lock of the document (step S21), and ends the deletion process. .
[0025]
FIG. 8 is a flowchart for explaining the search processing in the full-text search device according to the embodiment of the present invention.
To execute the search process, the user first inputs a search character string (search data) from the input means 1 (step S31). Next, in the search processing means 5, a token is obtained from the search character string using the text dividing means 6 (step S32). Then, all the tokens are S-locked by the lock processing unit 12 (step S33), and the search processing unit 5 uses the search large-scale full-text index in the search large-scale full-text index storage unit 8 to include a document including the search character string. A set (Rs) of document identifiers of data is obtained (step S34), and a set of document identifiers of document data including a search character string (using the registration small full-text index of the registration small full-text index storage unit 9) ( Ri) is obtained (step S35). Further, the search processing means 5 obtains a set (Rd) of document identifiers of the document data including the search character string using the delete small full-text index of the delete small full-text index storage unit 10 (step S36). Then, the lock processing means 12 unlocks all tokens (step S37), and the search processing means 5 performs the following set operation on the obtained set of document identifiers (Rs, Ri, Rd) (step S38). ), And the result is set as a search result (R), and a set of document identifiers of document data including the search character string is output to the user through the output means 2 (step S39). The S lock will be described later.
R = Rs + Ri-Rd
However, + is a logical sum operator and − is a logical difference operator.
[0026]
The search process will be described in detail using the full text index 63 of FIG. 5 as an example.
If the search character string is “full text search”, the text dividing unit extracts three tokens of “full text”, “sentence check”, and “search”. Next, the three transposed lists of the corresponding tokens in the full text index 63 are examined. When searching for a token having a difference in token appearance position of 1, a “full text search” exists in the 11th and 31st characters of the document identifier 1.
[0027]
FIG. 9 is a flowchart for explaining merge processing in the full-text search apparatus according to the embodiment of the present invention.
If you use a transposed list for the data used for the transfer process, you can use the transposed list that has already been created at the start of processing, compared to the case of registering / deleting using the original document data. Data transfer time can be shortened because time required for token extraction and transposition list creation by division processing is not required. In the present invention, the data transfer processing is called merge processing because it is processing between transposed lists.
[0028]
In order to execute the merge process, first, it is determined whether or not there is a token in the deletion small full-text index storage unit 10 (step S41), and if it exists (YES in step S41), all of the deletion small full-text indexes are determined. Processing of steps S43 to S45 is performed on the token. That is, in step S43, an X lock is applied to the tokens to be merged. In step S44, the transposed list of the token is extracted from the full-text index, and the appearance position information in the extracted transposed list is deleted from the transposed list of the corresponding token of the large-scale search full-text index (the transposed list is searched for a large-scale search). Delete from the full text index storage unit 8). In step S45, the X lock of the token to be merged is released, and the process returns to step S41. On the other hand, in the case of NO in step S41, the deletion small-scale full-text index storage unit 10 is emptied (step S42).
[0029]
Next, it is determined whether or not there is a token in the registration small full-text index storage unit 9 (step S46). If it exists (YES in step S46), the step is performed for all tokens of the registration small full-text index. The process of S48-S50 is performed. That is, in step S48, an X lock is applied to the tokens to be merged. In step S49, the transposed list of the token is extracted from the full-text index, and the previous transposed list is added to the end of the transposed list of the corresponding token of the large-scale search full-text index (the transposed list is added to the large-scale full-text index storage unit for search 8). To register). In step S50, the X lock of the merged token is released, and the process returns to step S46. On the other hand, if NO in step S46, the registration small full-text index storage unit 9 is emptied.
[0030]
FIG. 10 is a diagram for explaining the outline of the merge process, taking the transposed list of the token “full text” in the full text index 63 in FIG. 5 as an example.
The transposed list {1, 2, (11, 31)}, {2, 2, (30, 42)} for “full text” and the transposed list 72 of the full text index for deletion as the transposed list 71 of the full text index for search As a result of executing the merge process 73 with the transposed list {1, 2, (11, 31)} for “full text”, the transposed list {2, 2, (30, 42)} for “full text” (74 ) Is obtained. Further, this transposed list 74 and the transposed list {5, 2, (4, 16)}, {8, 1, (3)} for “full text” as the transposed list 76 of the registration full-text index are merged. 75, the transposed list {2, 2, (30, 42)}, {5, 2, (4, 16)}, {8, 1, (3)} (77) for “full text” is obtained. .
[0031]
(Merge processing mode 1)
The merge processing is started by the registration processing means 3 when the number of document identifiers registered in the registration small full-text index in the registration small full-text index storage unit 9 reaches a number designated in advance. It may be.
[0032]
(Merge processing mode 2)
The merge processing may be started by the registration processing means 3 when the storage capacity (size) in the small-scale full-text index storage unit 9 for registration reaches a size specified in advance. In this form, when used as an application form in which the size of document data registered by the user varies, when small document data is continuously registered, the registration to the small scale full-text index for registration is performed. It is possible to prevent the merge process from being started before the registration time becomes long. The merge processing time can be equalized by setting the size as a start condition. Further, in the case of the above-described merge processing (form 1), there is an advantage that the processing is simplified because the number of cases is used as a start condition and it is not necessary to manage the size of the full-text index storage unit.
In modes 1 and 2, the throughput of the entire update process can be increased even when sentences are frequently registered.
[0033]
(Merge processing mode 3)
The deletion processing unit 4 starts the merging process of the deletion small full-text index. The activation condition may be when the number of document identifiers registered in the small text index for deletion reaches a number specified in advance.
[0034]
(Merge processing mode 4)
The deletion processing unit 4 starts the merging process of the deletion small full-text index. The activation condition may be when the size of the small-scale full-text index storage unit 10 for deletion reaches a size specified in advance.
In modes 3 and 4, the throughput of the entire update process can be increased even when sentences are frequently deleted (delete process).
[0035]
With the above-described forms of merge processing, the full-text search apparatus can start full-text index merge processing under conditions suitable for the characteristics of document data to be registered / deleted and the characteristics of the field of use. The number of times can be reduced, and the throughput of the entire system can be improved.
[0036]
FIG. 11 is a flowchart for explaining the lock processing in the full-text search apparatus according to the embodiment of the present invention.
There are two types of lock modes, X and S. When an X lock is applied to an object, other users cannot lock the object. Also, if an S lock is applied to an object, other users can only apply an S lock to that object. The lock processing procedure performs exclusive control between objects by such a mechanism. In the lock processing in the lock processing means 12, first, when there is a lock request (YES in step S61), it is determined whether the lock is X lock (step S62). If YES in step S62, it is determined whether it is already locked (step S63), and if it is locked, it waits for the lock to be released (step S65). (Step S66), the process is terminated. On the other hand, if NO in step S62, it is determined whether it is already locked (step S64). If locked, the process proceeds to step S65, waits for the lock to be released, returns to step S62, and if not locked. S lock is applied (step S67), and the process is terminated.
[0037]
For example, when an X lock is applied to a token, a user who tries to apply an S lock to the token for searching waits until the X lock is released. In addition, if an S lock is applied to a token for a certain user to search, the merge processing procedure that attempts to apply an X lock to the token in order to perform the merge processing ends the search and the S of the token. Wait until the lock is released.
[0038]
As mentioned above, although each embodiment has been described centering on the full-text search device of the present invention, the present invention is a full-text search method including processing procedures in these full-text search devices, a program for functioning as these full-text search devices, Alternatively, the present invention can be implemented as a program for causing each unit to function, or as a computer-readable recording medium on which the program is recorded.
[0039]
An embodiment of a recording medium storing a program and data for realizing a full text search function according to the present invention will be described. As the recording medium, specifically, a CD-ROM, a magneto-optical disk, a DVD-ROM, an FD, a flash memory, and various other ROMs and RAMs can be assumed. This function is facilitated by causing a computer to execute the function of the device and recording and distributing a program for realizing the full-text search function. Then, the recording medium as described above is mounted on an information processing apparatus such as a computer and the program is read by the information processing apparatus, or the program is stored in a storage medium provided in the information processing apparatus. By reading, the full-text search function according to the present invention can be executed.
[0040]
【The invention's effect】
According to the present invention, since the registration / deletion process in the full-text search device is performed on a small-scale full-text index storage unit, the processing time can be shortened and the response time to the user can be shortened. It becomes possible. Furthermore, according to the present invention, when registering / deleting data in / from the full-text index for search, it is possible to directly use the transposed list already created while locking the token. The merge process time can be shortened, and the search process can be performed simultaneously.
[Brief description of the drawings]
FIG. 1 is a block diagram for explaining functions of a full-text search apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a hardware configuration example when the full-text search device in FIG. 1 is configured as a stand-alone.
FIG. 3 is a diagram illustrating a hardware configuration example when the full-text search apparatus in FIG. 1 is configured by a server / client.
FIG. 4 is a flowchart for explaining registration processing in the full-text search device according to the embodiment of the present invention.
FIG. 5 is a diagram for explaining processing in the full-text search apparatus of FIG. 1, and is a diagram illustrating an example of a full-text index.
FIG. 6 is a flowchart for explaining deletion processing in the full-text search apparatus according to the embodiment of the present invention.
FIG. 7 is a flowchart for explaining deletion processing in the full-text search device according to the embodiment of the present invention.
FIG. 8 is a flowchart for explaining search processing in the full-text search device according to the embodiment of the present invention.
FIG. 9 is a flowchart for explaining merge processing in the full-text search device according to the embodiment of the present invention.
FIG. 10 is a diagram for explaining the outline of the merging process taking the transposed list of the token “full text” of the full text index in FIG. 5 as an example;
FIG. 11 is a flowchart for explaining lock processing in the full-text search device according to the embodiment of the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Input means, 2 ... Output means, 3 ... Registration processing means, 4 ... Deletion processing means, 5 ... Search processing means, 6 ... Text division means, 7 ... Document data storage part, 8 ... Large-scale full-text index storage for search 9, small full-text index storage unit for registration, 10 small-scale full-text index storage unit for deletion, 11 merging unit, 12 lock processing unit, 21, 31 input device 22, 32 display unit, 23 , 33 ... I / O control device, 24, 34, 52 ... Main control device (CPU / memory), 25, 53 ... Storage device, 30 ... Client, 35, 51 ... Network control device, 40 ... Network, 50 ... Server.

Claims (5)

In a full-text search device that searches a document including a specified character string from a plurality of document data, a document data storage unit that stores registered document data, a full-text index storage unit for search, and data from a user are input Input means for outputting, output means for outputting search results, registration processing means for performing registration processing for document data, deletion processing means for performing deletion processing for document data, and search processing means for performing search processing, A full-text index storage unit for registration and a full-text index storage unit for deletion are provided separately from the full-text index storage unit for search, and the full-text index storage unit for registration and a full-text index storage for deletion from parts, the merge means for merging the data to full-text index storage unit for searching, and a locking means for locking process, the full text index storage unit and full-text index storage for deletion for the registration From the time of merging the data to full-text index storage unit for the search, said merging means performs a process for each inverted list of tokens which is a component of the full-text index, the token of the locking means said inverted list full text search device you characterized in that the Ru a lock. The merging means performs a process of merging data into the full-text index storage unit for search when the number of document data registered in the full-text index storage unit for registration reaches a predetermined number. The full-text search device according to claim 1 . The merger performs a process of merging data into the full-text index storage unit for search when the capacity of the full-text index storage unit for registration reaches a predetermined capacity. 1. The full-text search device according to 1 . The merging means performs a process of merging data into the full-text index storage unit for search when the number of document data registered in the full-text index storage unit for deletion reaches a predetermined number. the full-text search apparatus according to any one of claims 1 to 3, characterized. The merger performs a process of merging data into the search full-text index storage unit when the capacity of the full-text index storage unit for deletion reaches a predetermined capacity. The full-text search device according to any one of 1 to 3 .

Images