JPH08287105A - Document registration and retrieval device - Google Patents

Abstract

PURPOSE: To decrease file capacity and to speed up a retrieval process by compressing indexes by comparing the total number of pieces of information regarding a registered document with the total number of pieces of document information where a key appears, and expanding and using them at the time of the retrieval process. CONSTITUTION: When a document to be registered is inputted from a document input means 1, a key extracting means 2 extracts the key from the document by referring to a dictionary, etc., and an index generating means 11 generates an index by making the extracted key correspond to information regarding the registered document. An index compressing means 13 compares the total number of documents where the key appears with the total number of registered documents, compresses the document information included in the index in format which reduces the amount of data, and stores and holds it in file means 14 and 15. When the retrieval key is inputted from a retrieval key input means 16 at the time of the retrieval process, an index expanding means 13 expands the index stored in the file means prior to or simultaneously with the retrieval and a retrieval means 17 performs the retrieval process.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention registers a key extracted from a registered document and information about the document (for example, an identifier such as a document or a paragraph in the document) as an index, and the index corresponds to the search key. More specifically, the present invention relates to a document registration / retrieval apparatus that compresses and stores information about a document included in an index.

[0002]

2. Description of the Related Art Conventionally, in document registration and retrieval, a key extracted from a document to be registered and a key appearance such as the name of the document or a paragraph in the document, as described in Japanese Patent Laid-Open No. 2-186476, for example. The position information is associated with each other and registered as an index, and the search key input at the time of search is used to search the index for the corresponding document name or key appearance position in the document. As described above, the document registration search method using the index has an advantage that a predetermined document can be searched quickly at the time of search, and is widely and generally used.

Here, information about a document such as a registered document name and a key appearance position in the document is stored in the index. However, when the number of registrations increases, the index data amount increases and a large storage capacity is required. Therefore, it is required to make the index compact. Therefore, as described in Japanese Patent Laid-Open No. 5-257774, the index record number stored in the index in association with the key is compressed based on the statistical information to reduce the size of the file storing the index. Has proposed a document registration / retrieval device that expands a compressed index record number. In this document registration / retrieval device, the 1-byte difference method and the 2-byte difference method determined from the statistical information are used to store the 1-byte difference or 2-byte difference with respect to the reference data as an index record number compressed in the index.

[0004]

However, in the conventional document registration / retrieval device, since the process compresses the index based on the statistical information, the compression process becomes complicated and a lot of time is required for the process. Similarly, the decompression process becomes complicated and a lot of time is required for the process.
Therefore, the conventional document registration / retrieval device has a problem that even if the file capacity for storing the index is reduced, the registration / retrieval process is complicated and the rapid process cannot be performed.

The present invention has been made in view of the above-mentioned conventional circumstances, and stores an index while preventing a registration process and a search process from becoming complicated and delayed by using a simple and rational compression method. An object of the present invention is to provide a document registration / retrieval device that reduces the file size. In particular, according to the invention described in claim 1, the index compression and the corresponding decompression are performed in an optimum and simple format by comparing the total number of registered document information and the total number of document related keys. An object of the present invention is to provide a document registration / retrieval device. It is another object of the present invention to provide a document registration / retrieval device for compressing an index and decompressing the index in a rational format.

[0006]

In order to achieve the above object, a document registration / retrieval device according to a first aspect of the present invention includes a document input unit for reading a document to be registered and a key for extracting the input document. Key extraction means, index creation means for creating an index that associates information about a document for identifying a registered document with the extracted key, and a search key for searching a registered document In a document registration / retrieval device equipped with a search key input unit, a search unit for searching an index by using the search key, and an output unit for outputting a search result, in order to compress the index, An index compression unit that compresses information about the documents included in the index by changing the format based on the total number and the total number of documents in which the key appears, and a compression unit. A file means for storing the index, an index decompression means for decompressing the information about the document based on the corresponding compression format from the information stored in the file means in the search process, and an index for decompressing the information about the document. An index holding unit is further provided, and the index is compressed at the time of registration and expanded at the time of retrieval.

The document registration / retrieval device according to a second aspect is the document registration / retrieval device according to the first aspect, in which the index compressing means includes the total number of information regarding the registered documents and the information regarding the document in which the key appears. If the total number is equal to, the index is compressed in a format of deleting the information about the document.

The document registration / retrieval device according to a third aspect of the present invention is the document registration / retrieval device according to the first or second aspect, in which the index compression means displays the total number of information regarding the registered documents and the key. If the total number of information items related to documents for which the key did not appear is less than the total number of information items related to documents for which the key did not appear, use the information related to documents for which the key did not appear. The feature is that the index is compressed in a format of rewriting the index.

Further, a document registration / retrieval device according to a fourth aspect is the document registration / retrieval device according to any one of the first to third aspects, in which each bit of a bit string equal to the total number of information regarding the document is set. The information about the registered document is classified into a format indicating the information about each document by allocating the information about each document and inverting the bit and a format indicating the information about each document by a unique binary number. In the comparison between the total number and the total number of pieces of information regarding the document in which the key appears, one of the formats having a smaller amount of data is used to represent the information about the document of the index.

[0010]

In the document registration processing by the document registration / retrieval apparatus according to claim 1, when the document to be registered is input from the document input means, the key extraction means refers to the dictionary or the like provided in the apparatus from the document and the key is extracted. And the information about the registered document such as the document identification number and the extracted key are associated with each other, and the index creating means creates the index. Then, the index compressing unit compares the total number of registered documents with the total number of documents in which the key appears, and compresses the information about the documents included in the index in a format that requires a small amount of data. The file means stores and holds the index.

Further, in the document search process by the document registration / retrieval apparatus according to claim 1, when a search key is input from the search key input means, the search means searches the index by using the key and information about the corresponding document. The index decompression unit decompresses the index stored in the file unit prior to or at the same time as the search, and causes the retrieval unit to perform the retrieval process on the decompressed index. Then, the result obtained by the search is printed out by the output means and displayed on the display to be provided to the user.

According to another aspect of the document registration / retrieval device of the present invention, if the total number of information items regarding the document registered in the device is equal to the total number of information items regarding the document in which a certain key appears. , Delete information about the document for the key from the index. That is, when the information about all the documents registered in the device corresponds to a certain key, it is not necessary to identify the information about each document with respect to the key. Compress the index by not storing it.

According to another aspect of the document registration / retrieval device of the present invention, the index compression means compares the total number of pieces of information about the registered document registered in the apparatus with the total number of pieces of information about the document in which a certain key appears. In the case where the total number of pieces of information regarding the document in which the key does not appear is smaller than the total number of pieces of information regarding document in which the key appears, the index is rewritten with the information about the document in which the key does not appear. That is, the number obtained by subtracting the total number of pieces of information regarding the document in which a certain key appears from the total number of pieces of information regarding all the documents registered in the device (that is, the number of complement sets of the information regarding the document in which a certain key appears) is If the number is small, the index is compressed by using the small number of information.

According to another aspect of the document registration / retrieval device of the present invention, the index compression means compares the total number of information items regarding the document registered in the device with the total number of information items regarding the document in which a certain key appears. When the total number of information items regarding a document is small, the information about the document is identified and expressed by a unique binary code. When the total number of information items about a key is large, the information about the document is represented as a bit string. It is represented by.

For example, in the case where the information about 32 documents is identified and represented, the binary-coded decimal representation is 1
5 bits are required for each piece of information, and at least 5 × 32 is required to represent all 32 pieces of information while identifying them.
= 160 bits are required. On the other hand, in the same case, in the bit string display, if there are 32 bits, all information can be expressed while being identified. Comparing the two formats, when the total number of information to be expressed among the 32 pieces of information is 6, the binary coded decimal display requires at least 30 bits as a whole, and therefore the data amount is larger than that of the bit string display. If the total number of information to be represented is 7, the binary coded decimal display requires at least 35 bits, and the 32-bit bit string display requires a smaller amount of data. Thus, the index is compressed by representing the information about the document using whichever of the two formats is advantageous.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A document registration / retrieval device according to an embodiment of the present invention will be described below with reference to the drawings. As shown in FIG. 1, the document registration / retrieval device includes a document input unit 1, a key extraction unit 2,
Key list creating means 3, key list holding means 4, key list file access means 5, key list file 6,
File list creating means 7, file list holding means 8, file list access means 9, file list file 10, index creating means 11, index holding means 12, index file access means 1
3, index access file 14, index file 15, search key input means 16, search means 17,
The search result display means 18 is provided.

The document input means 1 is means for taking in the document as digitized data, and for example, an OCR device for reading a document on a paper medium, a device for reading an electronic information document from a storage means, an electronic information document It is composed of a communication device and the like for reading from the network. The key extracting means 2 is provided with a dictionary, analyzes the document read from the document inputting means 1 by referring to the dictionary, and extracts keys defined in the dictionary such as keywords and clauses from the document. This key extraction is performed using a known method such as morphological analysis or pattern matching.

The key list creating means 3 is the key extracting means 2
A key list, which is a list of keys that are extracted without duplication, is created and held in the key list holding means 4 composed of a readable / writable memory. The key list holding means 4 serves as a work area for other means in document registration processing and document search processing,
Holds a key list for the work. The key list file access means 5 reads and writes the key list between the key list holding means 4 and the key list file 6, and the key list held in the key list holding means 4 is used as the key list file in the document registration processing. Key list file 6 for document retrieval work.
The key list stored in is read and held in the key list holding means 4. The key list file 6 is composed of a hard disk device or the like, and stores the key list in a readable and writable manner.

The file list creating means 7 creates a file list which is a list of the document names and document identification numbers of the documents read from the document inputting means 1 without duplication, and is a file composed of a readable / writable memory. The list holding means 8 holds the list. The file list holding means 8 serves as a work area for other means in document registration processing and document search processing, and holds a file list for the work. The file list file access means 9 reads and writes the file list between the file list holding means 8 and the file list file 10, and the file list held in the file list holding means 8 is used as the file list file during the document registration processing. Stored in 10,
At the time of the document search operation, the file list stored in the file list file 10 is read and held in the file list holding means 8. File list file 10
Is composed of a hard disk device or the like and stores a file list in a readable and writable manner.

The index list creating means 11 creates an index in which the key extracted by the key extracting means 2 is associated with the information about the document from which the key is extracted, and the index holding means is a readable / writable memory. 12
To hold. In this embodiment, the document identification number is used as the information about the document, and the index stores the corresponding document identification number using the key as a headline as described later. The index holding means 12 serves as a work area for other means in document registration processing and document search processing, and holds an index for the work.

The index file access means 13 reads and writes the index between the index holding means 12 and the index access file 14 and the index file 15, and further performs compression processing and decompression processing. That is, the index file access unit 13 compresses the index held in the index holding unit 12 and stores it in the index access file 14 and the index file 15 in the document registration process, and the index access file 14 and the index access file 14 in the document search operation. Index file 1
The index stored in No. 5 is read out and expanded, and stored in the index storage unit 12. The index access file 14 and the index file 15 are composed of a hard disk device or the like, and store the contents of the compressed index in a readable and writable manner.

The search key input means 16 is a means for the user to input a search key for searching a document, and is composed of, for example, a keyboard. It is also possible to extract the search key by morphological analysis from a search request for a sentence or a phrase. In this case, the search key input means is, for example, an OCR device that reads a document in a paper medium, or an electronic information device. It comprises an apparatus for reading the document from the storage means, a communication apparatus for reading the electronic information document from the network, and the like, and has an analysis function.

The search means 17 uses the search key input from the search key input means 16 to generate the index holding means 1
2 is searched, the document identification number corresponding to the search key is extracted, and the file list held in the file list holding means 8 is searched using the document identification number, and the corresponding document name is searched. To extract. The search result display means 18 is means for outputting the document name searched by the search means 17 to the user, and is composed of, for example, a display device, a printer, or the like.

The index held by the index holding means 12 has the structure shown in FIG.
The index has a fixed part shown in (a) of the figure and a variable length part shown in (b) of the figure. The fixed part consists of the same number of items as the extracted keys, and the arrangement of these items is the same as the arrangement of the keys in the key list holding means 4. Therefore, each item corresponds to the extracted key.

Each item of the standard part includes the number of document files (document identification numbers) in which the key appears and a pointer to the document identification number list. Associated with the corresponding list of document identification numbers contained in. The variable length part includes, for each key, the identification numbers of the documents in which the key appears in a list, and each document identification number list stores all the document identification numbers in which the key appears in ascending order. Note that the pointer to the document identification number list is an absolute address or a relative address on the memory of the index holding unit 12, and other representation method may be used. That is, when the structure of the index is conceptually shown together with the fixed part and the variable length part, it becomes as shown in (c) of the figure, and the document identification number list corresponds to the number of keys and the same number of items. ing.

Index access file 1 described above
The content stored in No. 4 (hereinafter, simply referred to as an index access file) has the structure shown in FIG.
The index access file is composed of the same number of items as the extracted keys as shown in (a) of the figure, and the arrangement of these items is the same as the arrangement of keys in the key list holding means 4. Therefore, each item corresponds to each item of the index and each extracted key.

Each item of the index access file 14 includes the number of document files (document identification numbers) in which the key appears in correspondence with the index.
The format of the index and the address in the index file 15 are included. In the present embodiment, three bits a, b, and c are used as the index format as shown in FIG. 7B, and the combination of these bits compresses the document identification number stored in the index file 15. The format is shown.

In this embodiment, as shown in FIG.
Two compression formats are used. These formats are bit a,
It is shown as a combination of b and c. “001” in the sequence of bits a, b, and c indicates that “the document identification number is stored in the index file 15 because the key is included in all the registered documents (document identification numbers). “No” is indicated, indicating that compression is performed in a format in which the document identification number is not stored. The compression of this format is adopted because it is not necessary to identify the document registered with the document identification number for the key by the presence or absence of the key.

In addition, the sequence of bits a, b and c is "00".
"0" indicates that "the identification number of the document containing the key is stored as it is and stored in the index file 15", and only the identification number of the document containing the key is a binary code. This indicates that the storage format is being compressed. The compression of this format is adopted when the identification number of the document including the key is relatively small, which is less than the majority, compared with the total number of all registered document identification numbers.

In addition, the sequence of bits a, b and c is "01".
“0” indicates that “the identification number of the document that does not include the key among all the registered documents is represented by its numerical value and stored in the index file 15”,
This indicates that compression is achieved by storing only the identification number of the document that does not include the key as a binary code.
Note that the compression of the format is adopted when the identification number of the document including the key is more than the majority and is relatively large compared to the total number of all the registered document identification numbers,
All registered documents are identified for the presence or absence of a key by using a small document identification number that is a complement of the document identification number that includes the key.

In addition, the sequence of bits a, b and c is "10".
"0" indicates that "the identification number of the document that includes the key among all the registered documents is identified by the bit string notation and stored in the index file 15", and 2
It indicates that compression is performed in the form of identifying only the identification number of the document including the key by using a bit string having the same number as the total number of identification numbers of all the registered documents, instead of the evolutionary numerical value. The compression of the format is adopted when the total number of the identification numbers of the documents including the key is compared with the total number of all the registered document identification numbers, which will be described later.

The contents stored in the above-mentioned index file 15 (hereinafter, simply referred to as an index file) have the structure shown in FIG. The index file is made up of the same number of items as the extracted keys as shown in (a) of the figure, and the arrangement of these items is the same as the arrangement of keys in the key list holding means 4. Therefore, each item corresponds to each item of the index and the index access file, and each key corresponds to the extracted key.

In each item of the index file, as shown in (a) of the figure, the identification numbers of the documents in which the key appears in association with the index are collected in a list and stored as a document identification number list. The list of each item is compressed in the compression format shown by the index format, and the document identification number is stored in each list compressed in an appropriate format. That is, when the arrangement of the bits a, b and c is "001", the document identification number is not stored in the list for the key, and the document identification number list is empty.

In addition, the sequence of bits a, b and c is "00".
In the case of 0 ", as shown in (b) of the same figure, only the identification number of the document that is hit, that is, the document including the key is stored in the document identification number list as a binary number. Further, bits a and b are stored. , C in the arrangement of “010”, as shown in (c) of the figure, only the identification number of the document that does not hit, that is, does not include the key is 2 in the document identification number list.
It is stored as an evolved numerical value. Further, when the arrangement of bits a, b, and c is "100", as shown in (d) of FIG.
A bit string that identifies only the identification number of the document including the key is stored.

The above-mentioned index, index access file, and index file are collectively shown in FIGS. 5A, 5B, and 5C, respectively. Note that the total number of document identification numbers is 20, each identification number is a number from 1 to 20, and the total number of keys is 5. Each item of the fixed part in the index of (a) in the figure stores the total number of identification numbers of the documents corresponding to the respective keys, and a pointer to each document identification number list stored in the variable length part. Has been done. For example, the first item of the standard part has two document identification numbers corresponding to the key, and the document identification number list indicated by the pointer (ptr1) has document identification numbers "10" and "".
15 "is stored.

The index access file shown in FIG. 9B stores the total number of document identification numbers corresponding to the respective keys, the bit string indicating the compression format, and the address in the index file, as in the index. There is. Further, in the index file of (c) of the figure, the document identification number is compressed and stored in each of the areas having the address stored in the index access file as the start address.

For example, in the first item of the index access file, there are two document identification numbers corresponding to the key as in the index, and the address (addr)
In the area of the index file shown in 1), only the document identification numbers "10" and "15" including the key in correspondence with the compression format "000" are stored as binary numbers. The address addr1 is "0" because it is the head address of the entire area. In the second item of the index access file, the address column is empty because no document identification number is stored in the index file corresponding to the compression format “001”. In the third item of the index access file, the document identification number is stored in the index file as a bit string corresponding to the compression format “100”.

The selection criteria for the compression format of the above document identification number will be described. Here, let Li be the total number of identification numbers of the document including the i-th key, and if the document identification number is represented by a numerical value, D bytes (byte) per document identification number
If it is written in a bit string, K bytes are required. If the total number of registered document identification numbers is N, and the bit string is 8 bits and 1 byte, K is N / 8 bytes (N
Is divisible by 8), or K + 1 bytes (N is 8)
If not divisible by)).

First, when Li is equal to N, the compression format is "001", so no document identification number is stored in the index file. Then L
If i · D is less than or equal to K, the compression format is “000”, and as shown in FIG. 4 (b), the index file contains the identification number of the registered document including the key. Store with.

Next, when (N-Li) .D is less than or equal to K, it means that the compression format is "010", so that the index file contains a key as shown in FIG. 4 (c). Stores the identification number of a registered document that is not registered as a numerical value. Next, when Li · D exceeds K and (N-Li) · D also exceeds K, it means that the compression format is “100”.
As shown in FIG. 4D, the identification numbers of all registered documents are set as a bit string, the bits corresponding to the identification numbers of the documents including the key are turned ON, and the other bits are turned OFF and stored in the index file.

A procedure of main registration processing by the document registration / retrieval apparatus having the above configuration will be described with reference to FIG. First, when the document input unit 1 accesses the document to be registered and reads all the documents to be registered (step S1), a list of the document names of the documents read by the file list creating unit 7 is created,
Further, a unique document identification number is given to each of these document names to create a file list (step S2). The created file list is stored in the file list file 10 via the file list holding means 8 and the file list file access means 9.

After the file list is created or at the same time, the read document is analyzed by the key extracting means 2 in order to extract the key from each document (step S3), and the keys extracted by the key list creating means 3 are all the documents. In step S4, a key list is created as a list without duplication. The created key list is stored in the key list file 6 via the key list holding means 4 and the key list file access means 5.

After the key list is created or at the same time,
The index creating means 11 creates an index that is a list of keys and the document identification numbers of the documents in which the keys appear according to the order of the key list (step S5), and the created indexes are index holding means 12 and index file access means 13 Are stored in the index access file 15 and the index file 15 via the. The series of processes described above is repeated until there are no documents to be registered that have been read, and the process ends (step S6). In this embodiment, the document identification number starts from 1 in the order in which the documents are registered and continues to 2, 3, ..., And the document identification number of the last registered document is the same as the total number of registered documents. It becomes a value.

FIG. 7 shows the detailed procedure of the processing relating to the index file creation in step S5.
In the following example, the total number of identification numbers of all registered documents is N, the total number of extracted keys is M, the identification number of the registered document having the i-th key is Li, and a numerical value representing the document identification number is stored in the file. Let be the size of D bytes. It is also assumed that the index file access means 13 compresses the indexes shown in FIGS. 2A and 2B and stores them as the index access file and the index file shown in FIGS.

First, the size K of the area when the identification numbers of all registered documents are represented by bit strings is obtained (step S1).
1), create an index access file and an index file for each key. That is, the write start address in the index file and the total number Li of identification numbers of documents having the i-th key are written in the index access file (steps S12 and S13).
It should be noted that the write start address in the index file is 0 at the beginning and indicates the address to be written next when there is a write.

Next, it is checked whether or not the number Li of identification numbers of registered documents having the i-th key is equal to the total number N of identification numbers of all registered documents (step S14). Only the access file is written, and the document identification number is not written in the index file (step S15). That is, the "index format" of the index access file
Is set to "001". On the other hand, if Li and N are not equal, it is checked whether Li · D is K or less (step S
16), it is checked which is smaller, the case where the document identification number of the document having the i-th key is expressed by a numerical value or the case where it is expressed by a bit string.

As a result, when Li · D is K or less,
The identification number of the document having the i-th key is written numerically in the index file (step S17), and the index format of the index access file is "00".
By writing 0 "(step S18), it is indicated that the identification number of the document having the i-th key is written in the index file by a numerical value. On the other hand, when Li · D exceeds K, the i-th key is set. It is checked whether the total number (N-Li) · D of identification numbers of documents that the user does not have is K or less (step S
19) It is checked which of the case where the identification number of a document having no key is represented by a numerical value and the case where the identification number of a document having a key is represented by a bit string has a smaller area.

As a result, when (N-Li) .D is K or less, the document identification number of the document having no i-th key is written in the index file as a numerical value (step S2).
0), "010" is written in the index format of the index access file (step S21),
Indicates that the document identification number of a file that does not have a key is written as a numerical value in the index file. On the other hand, (N-
If Li) · D exceeds K, the presence or absence of the i-th key for the identification numbers of all registered documents is written in the index file as a bit string (step S2).
2), write "100" in the index format of the index access file (step S23),
Indicates that the identification number of the document having the i-th key is written in the index file as a bit string. Then, the series of processes described above is repeated M times, which is the total number of keys, to end the process (step S24).

The steps S14, S16,
The order of the determination process of S19 may be changed, and the determination process can be arbitrarily set in practice. Through the series of processes described above, for example, the index access file shown in FIG. 5B and the index file shown in FIG. 5C are created from the index shown in FIG. The document identification number is compressed and stored by a rational and simple method.

FIG. 8 shows an index access file and a procedure of a process of expanding the index file from the index file by the index file access unit 13 to reproduce the index, which is performed in the search process. In this process, the content of the fixed part of the index is written by the index access file and the information extracted from the index file, and the document identification number stored in the index file is decompressed based on the format of the index to change the variable length of the index. Write on the part. The area of the standard part of the index is, for example,
The total number M of keys is obtained from the key list file 6 and generated in the index holding means 12 in advance.

First, the total number Li of identification numbers of documents having the i-th key and the index format which is the compression format in the index file are read from the index access file (steps S31 and S32). Then, it is checked whether the index format is "001" (step S33), and if all the registered documents have the i-th key, the area of the list of the document identification numbers corresponding to the total number of the identification numbers of all the registered documents. Is stored in the index holding means 12, and a numerical value is written in the list so that the leading identification number is “1” and the last identification number is the total number “N” of the identification numbers of all registered documents in the area. The document identification number having the key (the number of registered document files) and the pointer to the document identification number list are written in the standard part of the index (step S3).
4).

On the other hand, if the index format is not "001" and all registered documents do not have the i-th key, the address in the index file is read from the index access file to obtain the write start address. The document identification number is prepared to be read from the index file (step S35). Next, it is checked whether the index format is "000", and it is checked whether the document identification number of the document having the i-th key is written as a numerical value (step S36), and the document identification number of the document having the i-th key. Is written as a numerical value, the index holding means 12 secures an area of the list of the document identification numbers for the total number Li of the identification numbers of the documents having the i-th key, and the numerical value of the document identification number is obtained from the index file. The data is read out and written in the document identification number list, and the total number Li of document identification numbers having the i-th key and the pointer to the document identification number list are written in the standard part of the index (step S37).

On the other hand, if the identification number of the document having the i-th key is not written numerically, it is checked whether the index format is "010" and the document having no i-th key is checked. It is checked whether or not the identification number of is written numerically (step S38). As a result, when the identification number of a document that does not have the i-th key is written as a numerical value, an area of the document identification number list for the total number Li of document identification numbers that have the i-th key is stored in the index holding means 12. Secure, read the numerical value of the document identification number from the index file, convert it to a bit string, invert "0" and "1" of each bit, then reconvert the bit string to a numerical value to create a document identification number list. Writing, the total number Li of the document identification numbers having the i-th key and the pointer to the document identification number list are written in the fixed part of the index (step S3).
9).

On the other hand, in cases other than the above, an area of the document identification number list for the total number Li of the identification numbers of the document having the i-th key is secured in the index holding means 12, and the bit string of the document identification number from the index file is secured. Is read, converted into a numerical value and written in the document identification number list, and i
The total number Li of document identification numbers having the th key and the pointer to the document identification number list are written in the fixed part of the index (step S40). The series of processes described above is repeated M times, which is the total number of keys, and the process is terminated (step S4).
1).

Note that the order of the judgment processing in steps S36 and S38 described above may be changed, and the processing is performed by judging the condition "0" or "1" bit by bit for each column of the index format. It may be changed, and the procedure can be arbitrarily set in practice. By the series of processes described above, for example, the index shown in (a) of FIG. 5 is created from the index access file shown in (b) of FIG. 5 and the index file shown in (c) of FIG. The document identification number is decompressed and stored in a rational and simple manner.

In the present invention, the index does not have to be divided into the fixed part and the variable length part, and these may be integrated. Further, the document identification numbers may be stored in the index in descending order instead of ascending order, and the document identification numbers stored in the index may be represented by bit strings instead of numerical values. Further, in the above-described embodiment, the example in which the document identification number is used as the information regarding the document is shown, but the document name, the physical address of the document file, or the like may be used.
A document such as a document identification number does not specify a document on a file-by-file basis, but may be a number, a symbol, or the like that specifies a chapter, paragraph, sentence, or the like in the document. In addition, not only the document name etc. is output for the information about the searched document,
You may also display the sentence, paragraph, etc. in the document which the search key hit.

The index access file does not have an index format indicating the compression method,
Each time the data is expanded and expanded in the memory, the number of documents having a key may be compared with the total number of registered documents to expand the document identification number. Further, the document identification number may be assigned according to the registration date and time of the document, and may not be a continuous value. Moreover, even if the index access file for storing the index and the index file are configured by separate storage devices,
The area may be divided into one storage device. Also,
The indexes are saved in these files at the end of the operation of the document registration / retrieval device, at the end of the registration processing of the document registration / retrieval device, after a certain time has elapsed from the start of the operation of the document registration / retrieval device, and from the start of the operation of the document registration / retrieval device. Various settings can be made each time a fixed time has elapsed.

[0058]

As described above, according to the document registration / retrieval apparatus of the first aspect, the format is changed by comparing the total number of registered document information and the total number of information regarding the document in which the key appears, and the index is changed. Is compressed and stored in a file, and the compressed index is decompressed and used for search processing, so the index can be compacted by a simple and appropriate compression processing, and the file capacity required for storage can be significantly increased. Further, a quick search process is realized by using the index reduced by the simple and quick decompression process.

Further, according to the document registration / retrieval device of any one of claims 2 to 4, in a rational format based on the relationship between the total number of information regarding registered documents and the total number of information regarding documents in which a key appears. The index compression process and the decompression process corresponding thereto can be performed, and a simple and quick registration search operation can be realized. In particular, if the information about the document is unified into numerical values or bit strings and is expanded into the index, a much quicker search process can be realized. It should be noted that when registering / retrieving a document in which each document has the same cliché as in the patent publication, the capacity may be half or less than that in the case where the document is not compressed, and a great effect can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a document registration / retrieval device according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram showing a structure of an index according to an embodiment of the present invention.

FIG. 3 is a conceptual diagram showing a structure of an index access file according to an embodiment of the present invention.

FIG. 4 is a conceptual diagram showing a structure of an index file according to an embodiment of the present invention.

FIG. 5 is a conceptual diagram showing structures of an index, an index access file, and an index file according to an embodiment of the present invention.

FIG. 6 is a flowchart showing a procedure of index creation processing according to an embodiment of the present invention.

FIG. 7 is a flowchart showing a procedure of index compression processing according to an embodiment of the present invention.

FIG. 8 is a flowchart showing a procedure of index decompression processing according to an embodiment of the present invention.

[Explanation of symbols]

1 ... Document Input Means, 2 ... Key Extraction Means, 11
... Index creating means, 12 ... Index holding means, 13 ... Index file access means (index decompression means, index compression means), 14 ... Index access file, 15
... Index file, 16 ... Search key input means, 17 ... Search means, 18 ... Search result display means,

 ─────────────────────────────────────────────────── ─── Continued Front Page (72) Akio Yamashita Akio Yamashita 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Kazuo Aihara Sakado, Takatsu-ku, Kawasaki City, Kanagawa Prefecture 3-2-1 KSP R & D Business Park Building in Fuji Xerox Co., Ltd. (72) Inventor Tatsuomi Kita 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa Prefecture KSP R & D Business Park Building in Fuji Xerox Co., Ltd. (72) Inventor Hiroshi Yamaguchi 3-2-1 Sakado, Takatsu-ku, Kawasaki-shi, Kanagawa KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Inventor Shinji Kawamoto 3-2-1, Sakado, Takatsu-ku, Kawasaki-shi, Kanagawa KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Inventor Naomi Hiraoka 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa Prefecture KSP R & D Business Park Building Fuji Xerox Co., Ltd.

Claims (4)

[Claims] 1. A document input unit for reading a document to be registered, a key extracting unit for extracting a key from the input document, information about a document for specifying a registered document, and the extracted key. An index creating means for creating an associated index, a search key input means for entering a search key for searching a registered document, a search means for searching an index using the search key, and a search result output. In order to compress the index, the document registration / retrieval device including an output unit changes the format of the information about the document included in the index based on the total number of registered documents and the total number of documents in which the key appears. Index compression means for compression, file means for storing the compressed index, and file means for storing during search processing. The index decompression unit that decompresses the information related to the document based on the corresponding compression format from the stored information, and the index holding unit that holds the index decompressed the information related to the document, compresses the index at the time of registration, A document registration / retrieval device that expands an index for retrieval. 2. The document registration / retrieval device according to claim 1, wherein the index compression means, if the total number of information regarding the registered document and the total number of information regarding the document in which the key appears are equal, A document registration / retrieval device characterized in that an index is compressed in the form of deleting. 3. The document registration / retrieval device according to claim 1 or 2, wherein the index compression means compares the total number of information regarding the registered document with the total number of information regarding the document in which the key appears. If the total number of information items related to documents for which the key does not appear is less than the total number of information items related to documents for which the key has appeared, compress the index in a format that rewrites the index using information about the document for which the key did not appear. Characteristic document registration and retrieval device. 4. The document registration / retrieval device according to claim 1, wherein the information about each document is assigned to each bit of the bit string equal to the total number of information about the document, and the bit is inverted. By doing so, the total number of information regarding the registered documents and the total number of information regarding the document in which the key appears in the format indicating the information regarding each document and the format indicating the information regarding each document by a unique binary number In the comparison, the document registration / retrieval device characterized in that the information about the document of the index is represented by one of the formats having a small amount of data.