JPH0721805B2 - Dictionary data search method - Google Patents

Description

The present invention relates to a method for searching data (character string) corresponding to a headword, such as a Kana-Kanji conversion dictionary and a language translation dictionary, and particularly to a large capacity. The present invention relates to a dictionary data search method suitable for searching dictionary data.

[Conventional technology]

As a conventional method for searching a dictionary data file, for example, there are methods described in JP-A-55-83962 and JP-A-56-38661. In these methods, the first or second character of the entry word is searched as the primary search target, and the dictionary body storing the third and subsequent characters is secondarily searched from the obtained address information. Is.

[Problems to be solved by the invention]

The above-mentioned conventional technique does not consider the dictionary data search method when a large-capacity dictionary data file is connected, and even if two or more characters of the entry words match, a large amount of dictionary data different from the third character onward is generated. Since the search is performed, there is a problem that the search time is increased.

An object of the present invention is to provide a dictionary data search method capable of searching a large-capacity dictionary data file at high speed and realizing with a small buffer memory.

[Means for solving problems]

To achieve the above object, the present invention provides a physical minimum unit (1) of a drive in which a large-capacity dictionary data file is stored.
(Sector) as a reference, create a sub index file that collects only the head entry words of each sector of the dictionary data file, and create a 1-sector master index that collects only the head entry words of each sector of the sub index. Search the large capacity dictionary data file.

At that time, the size of the required buffer memory is 1 sector when reading the master index which is 1 sector, and 1 sector when reading the corresponding sector (1 sector) of the sub index by the master index.
Further, when reading the sector in which the corresponding entry word of the large-capacity dictionary data file is stored by the sub-index, it becomes 1 sector, and the buffer memory of the size of 1 sector is prepared, and the large-capacity dictionary data file is created. You can search.

[Action]

 The operation of the present invention will be described below.

The master index file is read and the character string to be searched is n of the index word string of the master index file.
It is searched (first order) whether it is the same as the nth or is between the nth and the (n + 1) th. Next, the nth sector of the sub index file is read. It is searched (secondary) that the character string to be searched matches with the m-th index word string of the read sub index (1 sector) or is between the m-th value and the (m + 1) th value. Next, using the offset value k for the dictionary data file set for each sector of the sub-index, the (k + m) th sector of the dictionary data file is read. The headword of the read dictionary data file is compared with the character string to be searched, and the matching headword is searched (third).

With these operations, any character string to be searched can be searched for the target entry word by only accessing the master index file, the sub index file, and the dictionary data file three times each time. A large-capacity dictionary data file can be searched at high speed. The size of the buffer memory required at that time is 1 for reading the master index file (1 sector), the sub index file (1 sector that should be applicable), and the dictionary data file (1 sector that should be applicable). A buffer memory size of only one sector is sufficient by requiring sectors and rewriting the contents of the buffer memory each time and using them in common.

〔Example〕

An embodiment of the present invention will be described below with reference to FIGS. 1, 2, and 3. 2 and 3 show the structure of dictionary data, which is the main part of the present invention, and FIG. 1 is a block diagram of the device 10 for performing dictionary search. The device 10 includes an input / output device 19 for inputting a search code and outputting a search result, and an external storage device (first storage device) for storing the dictionary 1 and other data.
Storage device 14), a control circuit 12 for controlling the external storage device 14 according to a command from a CPU (central processing unit) 11, and a CPU 1
The internal memory (second storage device) 13 directly read / written from 1, the input / output device 19 and the control circuit 12 are connected to the internal memory 15
The CPU 11 controls a program stored in the program area 15 and executes a dictionary search. The internal memory 13 is divided into a program area 15, a work area 16, a master index file area 17, and a buffer area 18. The outline of the operation of the apparatus 10 will be described below. The word input from the input / output device 19 is taken into the CPU 11, and the CPU 11 searches the dictionary 1 based on the word. The result of the search is output by the CPU 11 to the input / output device 19, and the series of operations is completed.
The dictionary search, which is the main part of the present invention, will be further described. As shown in FIG. 3, the dictionary 1 is composed of a master index file 100, a sub index file 200, and a dictionary body (dictionary data file) 300. The configuration of the dictionary 1 will be described with reference to FIG. Dictionary body 300
Is composed of headwords arranged in the order of aiueo and dictionary contents corresponding to the headwords. That is, the dictionary contents 311a “a”, 311b corresponding to the headword 311 “a”
With “A”, 311c “Go” and the like as one record, a record of the subsequent heading 312 “Aa” and so on is followed in the same manner to form the dictionary body 300. Further, the dictionary main body 300 is divided into blocks each having a certain fixed length to form a first block 310, a second block 320 ... The sub-index file 200 is formed by collecting the headwords at the head of each block of the dictionary body 300. That is, the dictionary body 300
The first headword 311 “A” of the first block 310 of the sub-index file 200 becomes the first headword 211 “A” of the sub-index file 200, and the first headword 32 of the second block 320 of the dictionary body 300.
1 "Akatsuki" becomes the next entry word 212 "Akatsuki" of the sub index file 200. In the same manner, the sub-index file 200 is created by sequentially collecting the first entry words of each block of the dictionary body 300. At this time, the entry words of the sub-index file 200 are arranged in the order of aioi, like the entry words of the dictionary body 300. Similarly to the dictionary body 300, the sub index file 200 is also divided into blocks of a certain length, which are a first block 210, a second block 220 .... The master index file 100 is created by collecting the first entry words of each block of the sub index file 200. That is, the first entry word 211 “A” of the first block 210 of the sub index file 200 becomes the first entry word 101 “A” of the master index file 100,
The first entry word 221 “Kyokugay” of the second block of the sub index file 200 becomes the next entry word 102 “Kyokugay” of the master index file 100. The master index file 100 is formed by repeating this operation. The index words of the master index file 100 are also arranged in the order of aiueo. Master index file of dictionary 1 constructed by the above method
The headwords of 100 and the sub index file 200 have the following characteristics. The word x, which is in the order of Iueo in the order of In ≦ x <In + ₁ with respect to two adjacent headwords In and In + _{1 of the} master index file 100, is defined as two adjacent words in the nth block of the subindex file 200. There is a relation of headword Jnk ≦ x <Jnk + ₁ or Jnk ≦ x <In + ₁ , and the corresponding headword exists in the Sth block of the dictionary body 300. Here, if the number of index words in each block in the sub index file 200 is a _m , Is. Therefore, each headword of the master index file 100 is sequentially compared while being counted, and when x <Ip is stopped, the p- _1th block of the sub index file 200 is loaded into the buffer area 18 in the internal memory 13, and each Compare headwords sequentially while counting, and stop when x <Jq, Is calculated, and the S block of the dictionary main body 300 is stored in the buffer area 18
, And sequentially compare it with the headword to obtain the target dictionary contents. Since the master index file 100 is used for each search process, the master index file area 17 is provided in the internal memory 13 and loaded from the external storage device 14 in advance. As an example, the search procedure for the word "Agana" is shown. First, the master index file 100 headword 1
Compared with “a” in 01, it is in a later order, so headwords continue 10
2 Compared with “Kyokugay”, the search is completed in the master index file 100 because it is in the earlier order. Since there are two headwords counted during this time, 2-1 = 1 is obtained and the first block 210 of the sub index file 200 is loaded into the buffer area 18. Sub index file 200
The first headword 211 "A" to "Agana" is in a later order, so it is compared with the following headword 212 "Akatsuki", and it is in a later order than the following headword 213 "Aji". The search for the sub-index file 200 ends. The number of headwords counted during this period is three, and before the first block 210, the headwords are Since, The second block 320 of the dictionary main body 300 and the buffer area
Load to 18. Since the first headword 321 "Akatsuki" in the second block of the dictionary body 300 is different from the word "Agana", it is compared with the next headword 322 "Agana", and since they match, the content 322a "Buy, redemption" is obtained.

According to the present embodiment, only the portion used by the dictionary 1 is loaded into the internal memory 13, which is effective in improving the efficiency of the internal memory. Further, according to the present embodiment, by setting the size of the fixed length block to be small, the required buffer area 18
Can be made smaller, which has the effect of reducing the memory. Furthermore, in the present embodiment, the number of times a part of the dictionary 1 is loaded from the external storage device 14 for searching is always twice,
It is effective in stabilizing the search processing time and speeding up. In addition, in this embodiment, even if the dictionary 1 has a large capacity, the internal memory 13 used for the search can be made small, and the load from the external storage device 14 can be performed twice, so that a small electronic computer can be used. However, since the search can be performed at a sufficiently high speed, it is effective in reducing the price of the large-capacity dictionary search device.

A second embodiment will be described with reference to FIG. In the first embodiment, the master index file 100 used for each dictionary search is made resident by providing the master index file area 17 in the internal memory 13. In this embodiment, as shown in FIG. 4, the master index file 10
0 does not have a specific area and is used by being loaded into the buffer area 18 for each search.

According to this embodiment, the master index file area 17
Since it does not occupy, there is an effect of reducing the size of the memory.

The third embodiment will be described below. The sub index file 200 and the dictionary body 300 are each divided into blocks of a certain length. In the present embodiment, the length of this block is the minimum data length that can be handled by the external storage device 14 and the control circuit 12 or an adjustment multiple of the minimum data length.

According to the present embodiment, when a block having the sub index file 200 and the dictionary body 300 is loaded on the internal memory 13, the physical address in the external storage device 14 can be easily obtained, which has the effect of reducing the search time. . Further, according to the present embodiment, the external storage device 14 to the internal memory 13
Since there is no useless part in the data loaded up, there is an effect of saving memory, and the beginning and end of the block and the loaded data match, so search for the beginning of the block, transfer to fill the beginning, Is unnecessary,
This has the effect of reducing the search processing time.

A fourth embodiment will be described with reference to FIG. Dictionary body 300
Is very large and has a large number of blocks,
The sub-index file 200 will also be large and have many blocks. Therefore, the number of entry words included in the master index file 100 increases,
The size may correspond to multiple blocks in the block size of the sub index file 200. At this time,
A collection of the headwords at the beginning of each block of the subindex file 200 is referred to as a subindex file A (second subindex file) 400, and the subindex file A400 is further divided into blocks to obtain the headwords at the beginning of each block. Collect and master index files
Forming 100. If sub index file A400
If the number of blocks in the sub-index file B is large and therefore the collection of headwords at the beginning of each block is large, the sub-index file B is divided into blocks as the sub-index file B (third sub-index file), and the head of each block is divided. To form a master index file 100. The size of the master index file 100 is limited by repeating the above operation and providing a plurality of sub index files n (nth sub index file). A search method in the case where there are a plurality of sub-index files n will be described with reference to FIG. 5, in which the sub-index file n has two stages. First, the master index file 100 is searched, and the sub index file A4 is searched by the procedure shown in the first embodiment.
Lead a specific block from 00. Since the relationship of the sub index file A400 to the sub index file 200 is the same as the relationship of the master index file 100 to the sub index file A400, a specific block of the sub index file A400 is searched in the same way as the master index file 100, Similarly, a specific block is derived from the sub index file 200. Search the specific block of the sub index file 200 to guide the specific block of the dictionary main body 300, and the dictionary main body 3
The method of obtaining the record to be searched from the specific block of 00 is as described in the first embodiment. Further, even when the number of sub index files increases, the search operation can be realized by repeating the same processing.

According to the present embodiment, by increasing the number of stages of the sub-index file, it is possible to search a dictionary having a very large number of headwords with a small-capacity memory, which has the effect of saving memory. Further, according to the present embodiment, since the search operation can be realized by repeating the same search processing even when the number of stages of the sub index file is increased, there is an effect of simplifying the search processing. Further, according to the present embodiment, when the number of sub-index files increases, the number of headwords to be searched each time is limited to the size of the buffer area 18, although the number of times data is loaded from the external storage device 14 increases. Since it is possible to shorten the entire processing time, the search time can be shortened.

The fifth embodiment will be described below. At the beginning of each block of the sub-index file 200, the total number of entry words included in the block located before the entry word is recorded as an offset prior to the entry word. As shown in the first embodiment, in order to search for a specific block of the sub index 100 and to determine a specific block of the dictionary body 300 that follows, all of the sub index files 200 that are located before the block are searched. It is necessary to obtain a value obtained by subtracting 1 from the sum of the total number of index words included in the block and the number of index words counted as a result of the search. In the present embodiment, the block in the dictionary main body 300 that follows is specified by adding the counted value while sequentially searching for the entry word to the offset value at the beginning of the block.

According to this embodiment, since the block number can be easily obtained at the time of searching, the search time can be shortened. Further, according to the present embodiment, even if the dictionary 1 is changed and the number of headwords is changed, the same search processing procedure may be used, so that the expandability of the dictionary 1 is improved.

The sixth embodiment will be described below. In this embodiment, the first index word 101 of the master index file 100 is preceded by the logical or physical address data on the external storage device 14 at the head of the sub index file 200, and each sub index file 200 At the head of the block, the logical or physical address data on the external storage device 14 at the head of the block of the dictionary body 300 indicated by the headword of the head of each block is placed. In the dictionary search, the address data (each block or master index file 100
Number of headwords counted in the middle search (n-1) x
By adding the operation result of the block length l, the logical or physical address of the specific block in the sub index file 200 or the specific block in the dictionary body 300 actually loaded from the external storage device 14 can be obtained.

According to the present embodiment, since the logical or physical address on the external storage device 14 can be easily obtained, the search time can be shortened. Further, according to the present embodiment, the number of headwords can be changed without changing the search processing procedure, which has the effect of improving the expandability of the dictionary. Furthermore, according to this embodiment, the sub index file can be stored without changing the search processing procedure.
Since the arrangement of the 200 and the dictionary body 300 on the external storage device 14 can be changed, the memory efficiency of the external storage device 14 can be improved.

The seventh embodiment will be described below. The offset data described in the fifth and sixth embodiments is placed at the beginning of each block of the master index file 100 and the sub index file 200. Further, in this embodiment, data not used as offset data is set as a terminator (identification code) at the beginning of each block of the dictionary body 300. As described in the fourth embodiment,
A method for deriving a block in the sub index file 200 from the master index file 100, a method for deriving a block in the dictionary body 300 by searching for a block in the sub index file 200, and a method in which there are many sub index files The method of deriving a sub-index file of a lower level from a sub-index file is the same. Therefore, the operation of sequentially leading the lower index files is repeated until the beginning of the loaded block becomes the terminator, and when the terminator is detected, the search method of the dictionary body 300 is switched.

According to the present embodiment, the number of stages of the sub index file can be changed without changing the search processing procedure.
This has the effect of increasing the expandability of and generalizing the search processing.

The eighth embodiment will be described with reference to FIG. Dictionary body 300
Is composed of a record consisting of headwords and contents as a unit. A delimiter 314 indicating the end of the entry word between the entry word 311 and the content 311a, length data (relative address value) 315 indicating the distance to the next entry word, and a delimiter 316 indicating the end of the length data 315. Place and also content
A delimiter 317 indicating the delimiter between 311a and the content 311b and a delimiter 318 indicating the delimiter of the entire record are arranged.
In the search of the dictionary body 300, if the word and the headword match, the content is obtained from the data following the delimiter 316 of the length data 315, and if the word and the headword do not match, the length data 315 is used. Then, the head position of the next record is obtained, the next headword is compared with the word, and the matching headword is retrieved by sequentially repeating the following.

According to the present embodiment, the length of each record, the length of the entry word in the record, and the length of the contents can be made variable, so that the memory efficiency of the dictionary main body 300 is improved. Further, according to the present embodiment, when the headwords do not match, the next headword can be easily obtained, which has the effect of improving the search speed.

The ninth embodiment will be described with reference to FIG. A terminator (identification code) 215, which is a code that is not used as an entry word, is provided between the entry words included in the sub index file 200. When searching the sub-index file 200 for the next headword when sequentially comparing headwords to obtain headwords that are later than the word, the terminator 215 is searched to find the next headword. Similarly, with respect to the master index file 100 and the plurality of sub index files, the entry words can be separated by the terminator and the entry words can be obtained one after another.

According to the present embodiment, since the length of the headword can be made variable, there is an effect of improving the memory efficiency of the master index file 100 and the sub index file.

The tenth embodiment will be described below. Master index file 100, sub index file 200, dictionary body
If 300 headwords are katakana, the 8-bit = 1-byte code corresponding to katakana is usually 10100110
(=) To 11011110 (=), and the most significant bit is always "1". When the entry word is an alphanumeric character, the code corresponding to each character is 00100000 (=) to 01111111 (=) if it is an ASCII code. Therefore, instead of placing a special symbol at the delimiter of the entry word as in the eighth and ninth embodiments, the bit of the last character of the entry word can be manipulated to make the delimiter of the entry word. For example, if the entry word is composed of Katanaka, the most significant bit of each character is always "1", so the most significant bit of the last character of the entry word is " When you compare an entry word with a word, set the most significant bit of the last character of the entry word to "1" to restore the original Katakana code.
ASCII code, Shift JIS code, kuten code also take a fixed value that does not participate in word identification By inverting the bit to indicate the delimiter of the character string, the record in the dictionary body 300 and the record, the contents in the record The content can be separated.

According to the present embodiment, it is possible to add a delimiter mark to the character string itself and handle variable length data without adding a delimiter, which has the effect of improving memory efficiency.

The eleventh embodiment will be described below. Address data indicating a specific address on the external storage device 14 is stored in the content corresponding to the entry word of the dictionary body 300. At the location indicated by the address data, data ordered by entry words such as image information and audio information, and a large amount of data are stored. The data is loaded based on the address data which is the contents of the dictionary obtained by the search processing.

According to this embodiment, the dictionary body 300 is made small by placing a large amount of data outside the dictionary body 300, and the number of headwords contained in the block is large, so that the search time is shortened.

A twelfth embodiment will be described with reference to FIG. Dictionary body 300
The last headword 319 in the first block 310 of “Akas”
If the first entry word 321 of the second block 320 is “Akatsuki”, the entry word 212 ′ of the sub-index 200 corresponding to the second block 320 is “Akatsuki”.
(Identifiable word head). All the headwords included in the first block 310 of the dictionary body 300 are in front of the last headword 319, "Acass", and "Akatsu" is in a rearward order of "Acass" or the first heading of the second block 220. The search procedure may be the same as that already described because it is in the order preceding the word 221 “Akatsuki”. That is, the word I (n, m) <Ix <I (n +,) with respect to the last word I (n, m) of the nth block and the head word I (n +,) of the following (n +) th block. Ix can be used as a headword for the upper sub-index file or master index file 100.

According to the present embodiment, the size of the sub index file and the master index file 100 can be reduced by using the word having the shortest word length among the words that can be used as the entry word, which is effective in saving memory. Further, according to the present embodiment, it is possible to use a headword having a short word length to be compared at the time of searching, so that there is an effect of reducing the search time.

A thirteenth embodiment will be described with reference to FIG. Dictionary body 300
The headword at the beginning of each block has been deleted from the dictionary body 300 after becoming the headword of the sub-index file 200. In the search process, if there is a headword that matches the word to be searched during the search of the sub-index file 200, the search target is obtained by obtaining the dictionary content written at the beginning of the block of the dictionary body 300 corresponding to the headword. Get the contents of. Similarly, sub index file 20
0, or the headword at the beginning of each block of the second index file can be omitted.

According to the present embodiment, overlapping headwords can be omitted, which has the effect of saving memory. Further, according to the present embodiment, since the headword at the head of the block is directly referenced from the upper sub-index file or master index file, the search process is shortened and the average search time is shortened.

A fourteenth embodiment will be described with reference to FIG. The dictionary 1 is arranged on the external storage device 14 as shown in FIG. That is, the master index file 100 is located at the center of the entire dictionary 1, the sub-index file 200 is divided into blocks, and the dictionary main body 300 is arranged so that the corresponding blocks of the dictionary main body 300 indicated by the entry word in each block are gathered nearby. Are divided and placed. Therefore, the total sum of the distances of the addresses on the external storage device 14 from the master index file 100 to the respective blocks of the sub index file 200 becomes the minimum, and for each block of the sub index file 200, the heading included in each block The sum of the address gaps between the corresponding block of the dictionary body 300 indicated by the word is minimized.

According to this embodiment, the external storage device is used in the dictionary search process.
Since the data to be sequentially loaded from 14 is always in a close place on the external storage device 14, the access time of the external storage device 14 can be shortened, which has the effect of shortening the search time.

The fifteenth embodiment will be described below. When the search of the master index file 100 leads to a specific block of the sub-index file 200, the next dictionary body 30
It is possible to predict where in the dictionary body 300 the specific block of 0 is located. That is, the block of the dictionary body 300 indicated by the entry word included in the first block of the sub index file 200 is the first to nth blocks, and the block of the dictionary body indicated by the entry word included in the second block of the sub index file 200 is Block is n _{+ 1th
Assuming} that the second block of the sub-index file 200 is derived as a result of searching the master index file 100, the block of the dictionary body 300 to be derived next is the nth ₊₁ to It is one of the m-th blocks. Therefore, immediately after the specific block of the sub index file 200 is loaded from the external storage device 14, the dictionary main body 300 of the dictionary main body 300 is loaded into the external storage device 14. Control access to. External storage device 14 While accessing, the CPU 11 searches for a specific block in the sub index file 200 loaded previously.

According to this embodiment, since the search of the sub index file 200 and the access operation of the external storage device 14 are performed in parallel, the search processing time can be shortened.

[Effects of the Invention] According to the present invention, a large-capacity dictionary can be searched with a limited memory capacity as a block size, so that there is an effect of improving memory efficiency.

Further, according to the present invention, since a large-capacity dictionary can be searched using a small-sized electronic computer with a small memory, there is an effect of reducing the price of the dictionary search device.

Further, according to the present invention, a large number of headwords are collected and divided into blocks, and the headwords at the beginning of each block are collected to form a higher sub-index file or master index file. The headwords arranged in the file have less overlapping of characters even if the headwords are adjacent to each other, and the number of characters to be compared at the time of searching can be small, which has an effect of reducing the search time. For example, in the dictionary shown in FIG. 1, when searching for the word "Agana", the comparison with the index word in the master index file is "A".
And "ki" are all you need. In the subsequent sub-index file, the order relation can be discriminated by "A", "Aka", and "Abi", and a given heading is obtained by comparing "Aka" and "Agana" in the dictionary body. If the entry word in the dictionary body is searched from the beginning, the first two letters of all the entry words in the first block are inspected.

In addition, according to the present invention, any data having a specified order relationship or magnitude relationship can be used as a headword,
Katakana headwords (in order of Iueo), English headwords (AB
It can be used for C-order), number headwords (123-order), etc. With any character string to be searched, you can always search the master index file, sub-index file, and dictionary data file once for a total of three times to search for the target entry word, and a high-capacity dictionary at high speed. You can search data files.

Also, the size of the buffer memory required at that time is
Only one block is required only for reading the master index file (1 block), the sub index file (1 block that should be applicable), and the dictionary data file (1 block that should be applicable). By using the present invention, the present invention can be sufficiently realized with a small buffer memory size for one block.

If the size of 1 block is 2048 bytes and the average headword length is 5 bytes, the master index file of 1 block can manage the sub index file of 410 blocks. Furthermore, each block of the sub index file 410 of each dictionary data file
Blocks can be managed. That is, a master index file of 1 block can manage a large capacity dictionary data file of 168100 blocks (344 megabytes). If two blocks of buffer memory were prepared, it would be possible to manage a large-capacity dictionary data file of 671,000 blocks (2.7 GB) by the same calculation.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG.
FIG. 3 is an explanatory diagram for explaining a dictionary structure of one embodiment of the present invention, and FIG. 4 is a block diagram showing another embodiment of the present invention, FIG. 5, FIG. 6, FIG. 7, FIG. FIG. 8, FIG. 9 and FIG. 10 are explanatory views for explaining different embodiments of the present invention. 1 ... Dictionary, 10 ... Dictionary data retrieval device, 11 ... CPU
(Central processing unit), 13 ... Internal memory (second storage device), 14 ... External storage device (first storage device), 100 ...
… Master index file, 200 …… Sub index file, 300 …… Dictionary body (dictionary data file).

Claims (15)

[Claims] 1. A dictionary data retrieval method for retrieving dictionary data to obtain conversion data for a headword, wherein a dictionary data file consisting of dictionary data and a plurality of headwords is physically stored in a first storage means. A minimum access unit (sector) or a block unit having an integral multiple of the minimum access unit (sector) is formed, and a sub index file in which the head entry words of each block of the dictionary data file are collected is created in the block unit. A master index file is created in which the head index words of each block of the file are collected, the master index file is temporarily stored in the second storage means, and the index words in the master index file corresponding to the character string to be searched are searched. Search for the index word in the sub-index file for the entry word The block is temporarily stored in the second storage means, the index word in the stored block of the corresponding sub index file of the character string to be searched is searched, and the corresponding block of the dictionary data file for the index word is searched. A dictionary data search method characterized in that a character string to be searched is obtained by temporarily storing in the second storage means and searching the stored block data of the dictionary data file. 2. The dictionary data search method according to claim 1, wherein the master index file is resident in the second storage means. 3. The dictionary data search method according to claim 1, wherein the sub-index file is a second sub-index file in which only the head entry words of each block of the sub-index file are collected. Dictionary data retrieval method characterized by being composed of multiple sub-index files generated by various methods. 4. The dictionary data search method according to claim 1, wherein at the head of each block of the sub-index file, a head entry word of each of the blocks or a heading number of the sub-index file. A dictionary data retrieval method characterized by recording an offset value indicating whether it is a word. 5. The dictionary data search method according to claim 1, wherein the head entry word of each block is stored at the head of each block of the master index file and the sub index file. A dictionary data search method in which an address value indicating a corresponding physical address or logical address of a data file is recorded. 6. The dictionary data search method according to claim 1 or 3, wherein an identification code for distinguishing from an index file is recorded at the beginning of each block of the dictionary data file. Dictionary data search method. 7. The dictionary data search method according to claim 1, wherein each dictionary word in the dictionary data file is followed by a relative address value with respect to the next dictionary word. Data retrieval method. 8. A dictionary data search method according to claim 1, wherein an identification code for distinguishing from a next entry word is recorded at the end of each entry word of the sub index file and the master index file. A dictionary data search method characterized in that 9. The dictionary data search method according to claim 1, wherein the last character code of each headword in the sub index file and the master index file is distinguished from the next headword. A dictionary data retrieval method characterized in that fixed bits that do not change in each headword are recorded by being inverted. 10. The dictionary data search method according to claim 1, wherein as the data corresponding to the entry word of the dictionary data file, an address value indicating a physical address or a logical address of the first storage device. A dictionary data retrieval method characterized by recording columns. 11. The dictionary data search method according to claim 1, wherein each headword of the master index file and the sub-index file is composed only of a word head that can be distinguished from other headwords. Dictionary data retrieval method characterized by. 12. The dictionary data search method according to claim 1, wherein the head entry word of each block of the dictionary data file and the sub index file is omitted. 13. The dictionary data search method according to claim 1, wherein the master index file and the sub index file are stored in the first storage means. 14. The dictionary data search method according to claim 13, wherein the master index file is located at a block position that divides the dictionary data file into approximately two parts, and each block of the sub index file is defined as the dictionary data. A dictionary data search method characterized by arranging them in the vicinity of the corresponding block of the file. 15. The dictionary data search method according to claim 1 or 13, wherein the sub index file is located at a block position that bisects the sub index file immediately after accessing the master index file, and the sub index file. A dictionary data search method characterized in that, immediately after the access, the vicinity of the corresponding block of the dictionary data file is accessed (seeked) in advance.