JP2000029884A - Character code registration search device and character code registration search method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the storage capacity of a tri-array structure as much as possible by introducing a new data structure for which a double array structure which is a one-dimensional array is developed further and supplying a free registration part in respective character codes so as to be overlapped with each other on a CHECK array. SOLUTION: One of the values of a BASE array is a parallel movement amount as before and is applied to a KANJI (Chinese character) code (of low frequency) not used often (a). The other value is equivalent to the suffix I1 in the horizontal direction of a second BASE array. Then, for instance, in an area for expressing first and second bytes for constituting a 7-bit code, addresses are divided into three and classified into the groups d1, d2, d3 of three kinds (b). The CHECK array supplies the free registration part in the respective KANJI codes so as to be mutually overlapped on the CHECK array by supplying the parallel movement amounts d1, d2 and d3 even, for the character codes (slave nodes) continued to the same character (master node) (c).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a key search technique, and in particular, registers a character string such as a kanji code to be searched for a key in a double array structure which is a one-dimensional array as a data structure, and searches for the character string. And a character code registration search method.

[0002]

2. Description of the Related Art In recent years, with the spread of electronic mail and the like, the amount of electronic documents held by individuals has increased dramatically.
For example, there are many people who process e-mails in the number of hundreds to nearly 1,000 per day. Several MB a day, several hundred M per year
It is not uncommon for the document data to be B to GB.

When a large amount of data is handled, by compressing the amount of data by omitting redundant portions in the data, it becomes possible to reduce the storage capacity or to transmit the data at high speed. Universal encoding has been proposed as a method capable of compressing various data by one method. From the technical trends described above, compression technology has become an indispensable technology.

However, when attempting to compress a digitized document such as Japanese or Chinese on a word basis, it is quickly determined whether a character string input from the document is a word registered in a dictionary in advance. There is a need to. Further, since there are many words to be judged, this dictionary must be organized so that there is no useless area (so-called sparse area) as much as possible.

In the following, one word unit of data is called a symbol (symbol) or a character, following the name used in the information theory, and the data is connected by an arbitrary number of words. The symbol string,
Alternatively, it is referred to as a character string.

On the other hand, in the compression of a language code, a character string such as a word (key set (key set)) is converted into a double array structure (ie, a tri-array structure) which is a one-dimensional array as a data structure having as small a storage capacity as possible. It is an important issue to develop a search technology for storing and searching this tri-array structure at high speed.

In particular, a word (key set) as a search target
It is considered that a quasi-static key set whose key set is known in advance is registered in the trie array structure that configures the dictionary in which is stored. . Such a data structure is called a quasi-static data structure.

In view of such a technical background, Aoe et al. Have developed a character code character string registration search technique using a double array structure, which is a one-dimensional array as a data structure for pattern matching a plurality of keys at high speed. Proposal (Junichi Aoe: High-speed digital search algorithm with double array, IEICE (D), Vol. J71-D, No. 9,
p. 1592-1600 (1988)).

FIG. 8 is a diagram for explaining the prior art. FIG. 8A shows two one-dimensional arrays of a BASE array and a CHECK array having a double array structure, and B and B of these arrays.
FIG. 8 shows the contents of the ASE sequence and the CHECK sequence.
(B) shows baby #, batchor #, badg
FIG. 8C shows a trie structure (state transition diagram) in which er #, badge #, and jar # are stored.
FIG. 8 is a diagram for explaining a parent-child node relationship in a trie structure and a search operation using a double array.

In such a high-speed digital search technology using a double array, baby #, batchor #, ba #
FIG. 8A shows a double array corresponding to a trie structure in which dger #, badge #, and jar # are stored.
As shown in FIG.
Two two-dimensional arrays of the ECK array are prepared.

These BASE sequence contents and CHECK
Sequence contents are baby #, batchor #, bad
This corresponds to a trie structure (state transition diagram) in FIG. 8B in which ger #, badge #, and jar # are stored.

Here, the subscript CHEC on the BASE array
The subscripts on the K array are all the node indices (numbers 1 to 27 enclosed by white circles) in the trie structure shown in FIG.
(A node indicated by).

In the parent-child node relationship in the trie structure as shown in the left side of FIG. 8C, the index n of the parent node (the node indicated by n surrounded by white circles) corresponds to the subscript of the BASE array. I do.

On the other hand, a node index m of a child node (a node indicated by m surrounded by a white circle す る) corresponds to a subscript on the CHECK array.

When searching for a node index m = g (n, a) of a branch a (character "a") connected to a parent node n, first, BA
By looking at the sequence location where the subscript is n in the SE array, the sequence content BASE [n] is obtained. Here, the function g is a state transition function that defines a state transition for a key on the trie structure, and is called a goto function. At this time, if the array content BASE [n] specified by the subscript n is a numerical value d (array content BASE [n] = d), the numerical value d
Indicates a kind of origin parallel movement amount at the subscript m on the CHECK array.

On the CHECK array, the CH further moved by the code value of the character "a" from the position on the CHECK array shifted by the value of the subscript d (origin translation value) in the origin.
The subscript on the ECK array is m. The value (CHECK array [m]) held in the subscript m (= d + code value of character “a”) on the CHECK array is the value n of the node index of the parent node (that is, array content BASE [n]). When it matches, it is understood that the character “a” is registered under the node n.

In general, the time required for searching for a child node (matching time) in the trie structure is delayed according to the number of sibling nodes connected to the same parent node (ie, the number of branches from the node). The search time required for child nodes in the character code string registration search technology using a trie structure with an array can be a high-speed search that depends only on the key length, without depending on the number of sibling nodes It has been disclosed.

FIG. 9A shows the conventional technique of FIG.
To explain the operation of additionally registering idioms composed of character codes linked to the character "den" such as the characters "voltage", "electricity", "train", "denno", and "telephone" in the double array structure FIG. 9B is a diagram for explaining the operation of enlarging the double array structure, and FIG. 9C is a diagram for explaining the operation of FIG.
It is a figure for explaining operation of enlargement of a trie corresponding to a double arrangement structure of (b).

In such a conventional character code character string registration search technology, when a character code connected to "den" is additionally registered, each character code ("voltage", "electric", "train") connected to the character "den" is registered. "," Digital Cyber "," Phone ",…)
Maintain the relative positional relationship between the code values.

On the other hand, in the case of the CHECK arrangement, the portions marked with white circles have already been registered and filled, and the characters connected to "den" are not simultaneously located in the empty space of the CHECK arrangement.

Therefore, these characters "voltage", "electricity", "train", "electronic brain", "telephone",.
In order to register an empty space in the ECK array, both the BASE array and the CHECK array are extended.

That is, as shown in FIG. 9 (b), the minimum parallel movement such that these characters "voltage", "electricity", "train", "electronic brain", "telephone",. The amount d is calculated, and the code value d of the character “den” in the BASE array (= BASE array [n], n = g
Write this value at the position of (root, den).

On the other hand, the subscript values p,
q, r, s, and t are values obtained by adding the code values of the characters “pressure”, “ki”, “car”, “brain”, “talk”,. It is. That is, the subscript value p
= G (n, pressure), q = g (n, qi), r = g (n,
Car), s = g (n, brain), t = g (n, story).

Then, the subscript values p,
At a position of q, r, s, t, n which is a node index of “den” which is a parent node is written. That is, CHECK array [p] = CHECK array [q] = CHECK array [r]
= CHECK array [s] = CHECK array [t] = n. Such a node n, a node p, a node q, a node r, a node s, a node t, and a corresponding branch “den”,
FIG. 9C shows a trie structure formed by the branch “pressure”, the branch “ki”, the branch “car”, the branch “brain”, and the branch “talk”.

[0025]

The character code registered in the character code is less than 7,000 characters in both Japanese and Chinese at the first and second levels. A limited number of them are used to create idioms.
For example, the character code that produces more than 10 idioms in Chinese is about 500 characters or less.

In the trie structure, as the number of character codes connected to a certain character code increases, as shown in FIG. 9, in order to simultaneously register all character codes as keys in the empty space on the CHECK array, Must be increased.

Therefore, in such a conventional character code character string registration and search technology, as the number of character codes connected to a certain character code on the trie structure increases, as shown in FIG. In order to register codes in the empty space on the CHECK array at the same time, the CHECK array must be enlarged.

That is, unlike the alphabets (a to z), a character code has many characters, and when the character code is registered in the CHECK array while maintaining the relative positional relationship between the character codes connected to the certain character code, the CHECK array is There was a problem that it had to be enlarged. In addition,
There was also a problem that there were many gaps (sparse (sparse)) because registration could not be made and there was a possibility that the space would be empty.

An object of the present invention is to solve such a conventional problem. In particular, a double-array structure, which is a one-dimensional array as a conventional high-speed, low-capacity dictionary data structure, is provided. We propose a new data structure that further develops, and introduce a new data structure that is different from the conventional double array for frequent character codes,
Two types of values are registered in the BASE array, and one of the values is applied to a character code that is not used very often (infrequently) as a conventional translation amount, and the other value is used as a second BAS
The suffix of the second BASE array is divided into a predetermined type (specifically, three types) according to the code value of a character connected to the frequent character code by applying the suffix of the second array to any one of the suffixes of the E array. , By giving a unique translation amount, each character code is given an empty registration point so as to overlap each other on the CHECK array, and as a result,
Without expanding the CHECK array as much as possible, all character codes as keys are registered simultaneously in the empty space on the CHECK array, and without expanding the CHECK array as much as possible, the relative positional relationship of each character code connected to a certain character code can be registered. A character code registration / search apparatus and a character code registration / search method for registering a character code in a CHECK array while maintaining the same and further avoiding as many gaps (sparseness) as possible due to failure to register. It is an issue.

[0030]

According to the first aspect of the present invention, which is made by the present invention, there is provided a character code which is a key search target in a double array structure which is a one-dimensional array as a data structure. In a character code registration search device that registers a character string and searches for the character string, a translation amount d required to register a character of each character string to be a key search target
, A first array having a subscript of the index 106c at the beginning of each character string to be searched for the key, and an identification means for identifying a registered value in the first array.
Among the character strings shown in the first array, a second array in which information relating to a specific character connected to the beginning of the character string is registered, and a parallel movement registered in the first array and the second array. Key candidate point calculating means for calculating the sum of the quantity d and a value corresponding to a character connected to the end of the character string; and the index 106c of the beginning of the character string using the sum obtained by the key candidate point calculating means as a subscript And a second arrangement in which the character code registration and search are registered.

According to the first aspect of the present invention, a key search target is obtained as a new data structure obtained by further developing the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. Index 106c at the beginning of each character string
, A second array in which information on a specific character connected to the beginning of the character string in the character string indicated by the first array is registered, and a key search calculated by the translation amount calculating means. A character string using, as a subscript, the sum of the translation amount d required to register the characters of each target character string in the first array and the second array and a value corresponding to the character connected to the end of the character string Introducing a new data structure having a second array in which an index 106c at the beginning of the character string is registered makes it possible to provide an empty registration location for each character code so as to overlap each other on the CHECK array, As a result, CH
Without expanding the ECK array as much as possible, all character codes as keys can be registered in the empty space on the CHECK array at the same time, and without expanding the CHECK array as much as possible, the relative positional relationship of each character code linked to a certain character code can be registered. It is possible to register in the CHECK array while maintaining it, and furthermore, it is possible to avoid as much as possible a large gap (sparse) due to failure to register. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured,
The storage capacity of a tri-array structure in which a key is added and registered later to expand the memory can be reduced as much as possible.

According to a second aspect of the present invention, which is made by the present invention to solve the above-mentioned problem, the character code registration / search apparatus according to the first aspect creates a list of character codes frequently used for idioms. List means 101 for outputting a selected character code 103c selected from the character code list, and frequent character code selecting means 1 for outputting a frequency threshold value as to how many character codes to select.
04, a frequent character code storage means 103 for storing the frequent character code selected from the list means 101, and outputting the selected frequent character code 103a and the index 106c of the selected frequency character code, and a character code. This is a character code dictionary in which idioms are registered.The work is divided according to whether or not the character of interest is the beginning of a idiom based on the selected character code. Means 106 for outputting each class 106a, and a class 10 obtained by classifying character codes connected to the frequent character codes at the beginning of the word from the dictionary means 106.
BASE array means as the second array for storing the category storage means 108 for storing the character code 6a and the additional number 103b of the selected character code 101c, and for storing the same in the index 106c of the same character code on the BASE array 102
A code classifying unit 107 for classifying a second character code of the idiom with some bits in the character code of the second character code in order to classify characters connected to the frequently occurring character code at the beginning of the word; A minimum parallel movement amount such that a value obtained by adding an arbitrary parallel movement amount 109a to the code value of each character code of the same class 106a for each class 106a comes to an empty place on the CHECK array. The parallel movement amount calculating means 109 for calculating the parallel movement amount 109a and the parallel movement amount 109a from the parallel movement amount calculating means 109 are stored.
And a translation amount storage means 110 for storing the subscript position corresponding to the index at the beginning of the word, and for each of the classes, the translation amount input from the translation amount calculating means for the code value of each character code of the class. The added sum is used as a subscript value on the CHECk array, an index of the head of the character corresponding to the parent of the character code of the same class is registered at the position of the subscript, and the value of the sum is calculated as ((conventional head) + (character of interest)) ), The key candidate point calculating means 111 as an index of the next word prefix, and the code classifying means 10
7 and the listing means 101
The parallel movement amount 1 for each class output from the parallel movement amount storage means 110 based on the serial number 103b from
A second BA as the second array for storing 10a
SE array 105, the translation amount 111a from the translation amount storage unit 110, and the translation amount calculation unit 1
And a CHECK array means 112 as the first array for registering the internal setting value 109a in each character code of the class from 09 to the index 106c at the beginning of the CHECK array.

According to the invention described in claim 2, according to claim 1
In addition to the effects described in (1), the parallel movement amount from the parallel movement amount storage means 110 is a new data structure that is a further development of the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. CHECK array means 112 as a first array for registering 111a and the internal setting value 109a in each character code of the class from the translation amount calculating means 109 at the index 106c at the beginning of the CHECK array.
BASE array means 102 as a second array to be stored in index 106c of the same character code on the BASE array, and code values from code classifying means 107. 107
a second BASE array 105 as a second array for storing the parallel movement amount 110a for each class output from the parallel movement amount storage means 110 based on "a" and the serial number 103b from the list means 101. A new data structure is introduced, two types of values are registered in the BASE array, and one of the values is applied to a character code that is not used very often (infrequently) as a parallel translation amount, and the other value is used. The second
The subscript of the second BASE array is applied to the frequent character code as one of the subscripts of the BASE array, and the subscript of the second BASE array is divided into three types according to the code value of the character connected to the frequent character code.
By giving each translation amount independently, CHE
It becomes possible to give an empty registration part to each character code by overlapping each other on the CK array, and as a result, without expanding the CHECK array as much as possible, all character codes as keys can be used in the CHECK array. Can be registered at the same time, and without expanding the CHECK sequence as much as possible, it is possible to register in the CHECK sequence while maintaining the relative positional relationship of each character code connected to a certain character code. As a result, it is possible to avoid as many gaps (sparseness) as possible. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to a third aspect of the present invention, which is made by the present invention to solve the above problem, in the character code registration and search apparatus according to the first aspect, at first, the root of the trie structure is designated as a word prefix, A document input unit 201 for setting an end mark # as an end symbol at the beginning of a word W, subsequently instructing input of a character code b as a character to be searched, and detecting the beginning of the inputted character code b. BASE array means 102 for inputting a numerical value 102a from the position corresponding to the initial W or the character code index 106c
And the numerical value 10 inputted from the BASE array means 102
It is determined whether 2a is an additional number of a frequently occurring initial character code, an index of an initial character code that is not so, or an index 106c of an initial W in the middle of the character string. Is output as an additional number 202a of the frequently appearing character code, and the numerical value 102a input from the BASE array means 102 is not the index 106c of the initial W and the frequently appearing character is output. A registered value judging means 202 which outputs a parallel movement amount 202b when the code is not a code; and a serial number of an initial character code in which the numerical value input from the BASE array means frequently appears in order to classify characters connected to the frequently occurring character code at the beginning of the word. Is BA
When input from the SE array unit, a code classifying unit 107 that classifies by several bits in the character code of the character code connected to the leading kanji, and a serial number of the frequent character code output from the registered value determining unit 202 202a, a second BASE array 10 for storing the translation amount 105a from a location corresponding to the classification of the code value of the character code b.
5 and the numerical value 1 input from the BASE arrangement means 102
Indices of initial character codes where 02a is not frequent or initial W
And the parallel movement amount storage means 110 for storing the numerical value 102a as the parallel movement amount when the index 106c is the index 106c. Is added as a subscript value on the CHECk array, the index of the prefix corresponding to the parent of the character code of the same class is registered at the subscript position, and the value of the sum is ((conventional prefix) + ( Key candidate point calculating means 111 as an index of the next beginning of word consisting of the character of interest))
CHECK arrangement means 112 for inputting a key to a place corresponding to sum 111a from key candidate point calculation means 111
And the key input by the CHECK arrangement means 112,
It is determined whether the index is equal to the index of the initial letter code or the index 106c of the initial letter W. If it is determined that the input key is equal to the index of the initial letter code or the index 106c of the initial letter W, the idiom is registered in the dictionary. A character code registration and search device having a key / prefix comparison unit 203 for determining that

According to the invention described in claim 3, according to claim 1
In addition to the effects described in (1), as a new data structure which is a further development of the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, the sum 111a from the key candidate point calculation means 111 CHECK arranging means 112 for inputting a key at a corresponding position, BASE arranging means 102 as a second array for inputting a numerical value 102a from a position corresponding to an index 106c of a letter W or a character code,
A second BASE array 105 as a second array for storing the parallel movement amount 105a from a location corresponding to the classification of the code value of the character code b as the additional number 202a of the frequently appearing character code output from the registered value determination unit 202 By introducing a new data structure having a key set, a dictionary is stored in which a quasi-static key set whose key set is known in advance is stored as a search target, and a key is registered and expanded later as needed. The storage capacity of a simple tri-array structure can be minimized. As a result, the data is stored in a double array structure (that is, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible, and the tri-array structure can be subjected to high-speed pattern matching using a search key. become.

According to a fourth aspect of the present invention to solve the above-mentioned problem, a character code character string to be searched for a key is registered in a double array structure which is a one-dimensional array as a data structure. In the character code registration search method to be searched, a parallel movement amount d calculating step of calculating a parallel movement amount d required to register a character of each character string as a key search object, and each character string as the key search object , A first arranging step using the index 106c at the beginning of the subscript as a suffix, an identifying step of identifying a registered value in the first arranging step, and a character string of the A second arrangement step in which information relating to a specific character connected to the beginning of the word is registered, the translation amount d registered in the first arrangement step and the second arrangement step, and a value corresponding to the character connected to the end of the character string Calculate the sum with A candidate point calculating step, a character code register searching method and a second sequence step of registering the prefix indication 106c of the character string sum obtained in the key candidate point calculation step as a subscript.

According to the fourth aspect of the present invention, a key search is performed as a new data structure which is a further development of the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. Index 106c at the beginning of each character string
A first arranging step having a subscript as a suffix, a second arranging step in which information on a specific character connected to the beginning of the character string in the character string indicated in the first arranging step is registered, The sum of the translation amount d required to register the characters of each character string to be searched for the key in the first arrangement step and the second arrangement step and the value corresponding to the character connected to the end of the character string And the second arrangement step in which the index 106c at the beginning of the character string is registered is used to introduce a new data structure to be formed. Can be given, and as a result, all character codes as keys can be changed to CH
Can be simultaneously registered in the space on the ECK array, and CHECK
CHEC while maintaining the relative positional relationship of each character code connected to a certain character code without expanding the array as much as possible
It becomes possible to register in the K array, and in addition, it is possible to avoid as many gaps (sparseness) as possible without registering as much as possible. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to a fifth aspect of the present invention, which is made by the present invention to solve the above problem, in the character code registration and search method according to the fourth aspect, the identification step includes the first step.
A step of identifying the registered contents registered in the arrangement step as one of the serial number I of the specific character code located at the head of the idiom, another character code at the head, or the translation amount d of the head of the character string And if the registration content registered in the first arrangement step is identified as the additional number I of the head specific character code, the additional number I in the second arrangement step
And obtaining a translation amount d by referring to the array location indicated by the character code registration search method.

According to the invention set forth in claim 5, according to claim 4,
In addition to the effects described in (1), as a new data structure that is a further development of the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, the prefix of each character string to be searched for a key is A first arrangement step using the index 106c as a suffix, a second arrangement step in which information relating to a specific character connected to the beginning of the character string in the character string indicated in the first arrangement step is registered, and a translation amount calculating means. Is the sum of the translation amount d required to register the characters of each character string to be searched for the key in the first and second arrangement steps and the value corresponding to the character connected to the end of the character string. A new data structure formed by using the second arraying step in which the index 106c of the beginning of the character string is registered by using as a subscript, and the registered content registered in the first arraying step is added to the beginning of the idiom. The specific character code Serial number I, the other at the beginning of the character code,
Or, if the registered content registered in the first arrangement step is identified as the additional number I of the first specific character code, the parallel movement amount d is identified as one of the parallel movement amounts d of the beginning of the character string. By executing the identification step of obtaining the parallel movement amount d with reference to the array position indicated by the serial number I, the CHE
It is possible to provide an empty registration location for each character code by overlapping each other on the CK array. As a result, all character codes as keys can be replaced on the CHECK array without expanding the CHECK array as much as possible. It is possible to register in a vacant space at the same time, and it is possible to register in the CHECK array while maintaining the relative positional relationship of each character code connected to a certain character code without expanding the CHECK array as much as possible. It is possible to avoid as many gaps (sparseness) as possible. Thus, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and a storage capacity of a tri-array structure in which a key is registered and expanded later as appropriate is minimized. CHEs for each character code
Compared to the case of giving the amount of translation on the CK array,
An increase in the K array can be suppressed, and the spatial efficiency of the double array can be improved.

According to a sixth aspect of the present invention, which is made by the present invention to solve the above-mentioned problem, in the character code registration and search method according to the fifth aspect, the additional number I of the head specific character code of the character string is provided. And a step of referring to the second arrangement step based on a classification of a character code connected to the leading specific character code.

According to the invention described in claim 6, according to claim 5,
The same effect as the effect described in (1) is obtained.

According to a seventh aspect of the present invention, which is made by the present invention to solve the above-mentioned problem, in the character code registration search method according to the sixth aspect, the character code registration search method is linked to the first specific character code in the second arrangement step. The step of classifying character codes is a character code registration search method using the code value of the character code.

According to the invention of claim 7, according to claim 5,
The same effect as the effect described in (1) is obtained.

According to an eighth aspect of the present invention, which has been made to solve the above-mentioned problem, in the character code registration / search method according to the fourth aspect, the second arrangement step is frequently performed in forming idioms. Is a character code registration search method including a step of selecting a character used for a character string as a specific character connected to the beginning of the character string.

According to the invention described in claim 8, according to claim 5,
The same effect as the effect described in (1) is obtained.

According to a ninth aspect of the present invention to solve the above-mentioned problem, in the character code registration search method according to the fourth aspect, the second arrangement step uses the same character as a idiom. In this case, there is provided a character code registration search method including a step of selecting, as a specific character connected to the beginning of the character string, a character string whose code value width is equal to or greater than a predetermined threshold value.

According to the ninth aspect of the present invention, the fifth aspect
The same effect as the effect described in (1) is obtained.

According to a tenth aspect of the present invention, which has been made to solve the above problem, in the character code registration and search method according to the fourth aspect, a list of character codes frequently used for idioms is created. And outputs the selected character code 103c selected from the list of character codes.
A list process (step S2) for selecting a prefix in a character code of a parent node to which character codes of a predetermined type or more are connected as child nodes and adding a new additional number 103b to the list and outputting the list; Frequent character code selecting step (step S2) for outputting a frequency threshold value regarding whether to select the character code of (1) and the list step (step S2)
A frequently-used character code storing step (step S2) of storing the frequently-used character code selected from the above and outputting the selected frequently-used character code 103a and the index 106c of the selected frequency character code; In the character code dictionary, the work is divided depending on whether or not the character of interest is the beginning of an idiom based on the selected character code. Dictionary process to output (step S3)
And a classification result storing step (step S3) for storing each class 106a obtained by classifying the character codes connected to the frequent character codes at the beginning of the word generated by the dictionary step (step S3).
9) and calculating the additional number 103b of the selected character code 101c, and storing it in the index 106c of the same character code in the BASE array, as the second array process (step S6, 11). ), And a code classification step of classifying the second character code of the idiom with some bits in the character code of the second character code in order to classify characters connected to the frequently appearing character code at the beginning of the word (step S7). ) And a value obtained by adding an arbitrary parallel movement amount 109a to the code value of each character code of the class 106a for each class 106a. Of parallel movement 109a
For calculating the amount of parallel movement (steps S8, S10)
And the parallel movement amount 109a generated by the parallel movement amount calculation step (steps S8 and S10) is stored, and the parallel movement amount 110a is stored in a subscript position corresponding to the index of the head of the BASE array means 102. A movement amount storing step (steps S8, S10, S11) and, for each class, the sum of the parallel movement amount input from the parallel movement amount calculation means added to the code value of each character code of the same class in the CHECk array. Is registered as a subscript value, and the index of the head of the character which is the parent of the character code of the same class is registered at the position of the subscript, and the sum value is set to the value of the next prefix consisting of ((conventional prefix) + (character of interest)) Key candidate point calculation step for use as an index (steps S9, 1
2), the code value 107a generated by the code classification step (step S7) and the list step (step S7).
The second arrangement step for storing the parallel movement amount 110a for each class output by the parallel movement amount storage step (steps S8, S10, S11) based on the additional number 103b generated in 2). A second BASE arraying step (step S9), and the parallel movement amount storing step (step S9).
8, 10, 11) and the internal setting value 109 in each character code of the class generated by the parallel movement amount calculating step (steps S8, 10).
A character code registration and search method including a CHECK arrangement step (step S12) as the first arrangement step of registering a at the index 106c at the beginning of the CHECK arrangement.

According to the tenth aspect, in addition to the effect of the fourth aspect, the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, is further developed. As a new data structure, the parallel movement amount 111a and the parallel movement amount calculation step (steps S8, 1) generated by the parallel movement amount storage step (steps S8, S11) are used.
0) generates the minimum translation amount 109a by CHECK.
A CHECK arrangement step (step S12) as a first arrangement step for registration at the position of the index 106c at the beginning of the arrangement;
The additional number 103b of the selected character code 101c is calculated, and the index 10 of the same character code on the BASE array is calculated.
6c, a BASE arrangement step (steps S6, 11) as a second arrangement step, and a parallel movement amount 110a for each class output by the parallel movement amount storage step (steps S8, 10, 11). The second as the second arrangement step
A new data structure created by executing the BASE array process (step S9) is introduced, and two types of values are registered in the BASE array, and one of the values is not used much as the conventional parallel movement amount ( The second value is applied to the frequent character code, and the other value is applied to the frequent character code as one of the subscripts of the second BASE array.
The subscripts of the array are divided into three types according to the code values of the characters connected to the frequently appearing character codes, and by giving their own parallel movement amounts, the registered locations that are vacant in each character code are overlapped on the CHECK array. As a result, all character codes as keys can be changed to CHECK without expanding the CHECK array as much as possible.
It becomes possible to register at the same time in the space on the array, and CH
Without expanding the ECK array as much as possible, maintain the relative positional relationship of each character code connected to a certain character code
It becomes possible to register in the HECK array, and in addition, it is possible to avoid as many gaps (sparseness) as possible without registering as much as possible. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to an eleventh aspect of the present invention for solving the above-mentioned problem, in the character code registration and search method according to the fourth aspect, a root of a trie structure is first specified as a word prefix, A document input step of setting an end mark # as an end symbol at the beginning W, instructing input of a character code b as a search target character, and detecting the beginning W of the input character code b ( Steps P2 and P3) and index 1 of the initial W or character code
BA to input numerical value 102a from the position corresponding to 06c
Whether the numerical value 102a generated by the SE arrangement step (step P4) and the BASE arrangement step (step P4) as the second arrangement step is an additional number of the frequently occurring initial character code,
It is determined whether it is an index of the initial character code that is not so, or it is an index 106c of the initial W in the middle of the character string, and when an index exceeding the range of the index constituting the trie is given, this index is determined. , Frequently added character code additional number 202a
And the numerical value 102a generated by the BASE arrangement step (step P4) is the index 1 of the prefix W.
Registered value judging step of outputting parallel movement amount 202b when character code is not 06c and not frequently appearing character code (step P5)
In order to classify characters connected to the frequently appearing character code at the beginning of the word, if the serial number of the initial character code in which the numerical value inputted from the BASE arranging step 102 (step P4) frequently appears is inputted from the BASE arranging means, A code classification step (step P6) of classifying the character codes of the character codes connected to the kanji by some bits, and an additional number 202 of the frequent character code generated by the registration value determination step (step P5)
a, a second BASE arrangement step as the second arrangement step for storing the translation amount 105a from a position corresponding to the classification of the code value of the character code b (step P7)
And when the numerical value 102a generated by the BASE arrangement step (step P4) is the index of the initial character code or the index 106c of the initial W, which is not frequently used, stores the numerical value 102a as the translation amount. Steps P7 and P8) and, for each class, the sum of the translation amount input from the translation amount calculating means and the code value of each character code of the class is used as a subscript value on the CHECk array. Register the index of the prefix corresponding to the parent of the character code of the same class in the subscript position, and change the value of the sum to ((conventional prefix) +
(Step P9), which is the index of the next word consisting of (character of interest)), and inputting a key to a place corresponding to the sum 111a from the key candidate point calculation step (step P9). CHECK as first arrangement step
The key generated by the arrangement step (Step P9) and the CHECK arrangement step (Step P9) as the first arrangement step is the first letter code index or the first letter W index 106c.
Key / initial matching step (step P10, step P10) in which if the input key is determined to be equal to the index of the initial character code or the index 106c of the initial W, it is determined that the idiom is registered in the dictionary. p11, P12).

According to the eleventh aspect of the present invention, in addition to the effect of the fourth aspect, the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, is further developed. As a new data structure, a CHECK arrangement step (step P9) as a first arrangement step of inputting a key to a place corresponding to the sum 111a from the key candidate point calculation step (step P9), and a prefix W or a character code index BASE arranging step (step P) as a second arranging step of inputting a numerical value 102a from a position corresponding to 106c
4) and a second array for storing the parallel movement amount 105a from a location corresponding to the classification of the code value of the character code b as the serial number 202a of the frequent character code generated in the registered value determination step (step P5) By introducing a new data structure having a second BASE array process (step P7) as a process, a dictionary is stored in which a quasi-static key set whose key set is known in advance is stored as a search target. Thus, the storage capacity of a tri-array structure in which a key is additionally registered as appropriate and expanded can be reduced as much as possible. As a result, the data is stored in a double array structure (that is, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible, and the tri-array structure can be subjected to high-speed pattern matching using a search key. become.

According to the twelfth aspect of the present invention, which has been made by the present invention to solve the above-mentioned problems, the tenth or eleventh aspects of the present invention are described.
In the character code registration search method described in, when the specific character connected to the beginning of the character string is an idiom, the width of the code value of the connected character code is determined as a specific character connected to the beginning of the character string. A character code registration search method including a step of selecting a character as described above.

According to the twelfth aspect of the present invention, in addition to the effects of the tenth or eleventh aspects, more vacant portions not filled by the conventional method are filled, and the BASE arrangement step ( B created in step P4)
The increase in both the ASE sequence and the CHECK sequence created in the CHECK sequence process (step P9) is also appropriately suppressed. In addition, since the number of times of processing refers to the second BASE arranging step (step P7), the number of times of processing is increased only once, so that the number of times of processing is almost the same.

[0054]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the following embodiments, a description will be given using a kanji code as a character code.
The character code registration and search device and the character code registration and search method of the present invention can be applied to a 2-byte character code used in a language system such as expressing a word or an idiom by combining a plurality of single characters having 00 types, More specifically, a JIS code in Japanese that can be expressed in a 7-bit code area (FIG. 7A), a 7-bit code in Chinese GB2312-80, or an 8-bit code area (FIG. 7B)
Area of Japanese EUC code and Chinese G
An 8-bit code of B2312-80 is conceivable.

The number of kanji registered in the kanji code is less than 7,000 in both Japanese and Chinese at the first and second levels. A limited number of them are used to create idioms. For example, the number of Chinese character codes that produce more than 10 idioms in Chinese is about 500 characters or less.

In the trie structure, as the number of kanji codes connected to a certain kanji code increases, as shown in FIG. 9, in order to simultaneously register all character codes as keys in an empty space on the CHECK array, Must be increased.

Therefore, this time, a double array (dou), which is a one-dimensional array as a conventional high-speed, low-capacity dictionary data structure, is used.
It is the present invention that has been devised in a further developed form of the ble-array structure.

Therefore, for these frequent kanji codes, a new data structure different from the conventional double arrangement is introduced.

FIG. 1 shows a new data structure of the present invention.
FIG. 1A shows a translation amount applied to a kanji code that is not used very often (infrequently) in a new data structure applied in the character code registration search device and the character code registration search method of the present invention. FIG. 1B shows the basic structure of a BASE array in which the frequent or additional numbers are registered.
In the new data structure of the present invention, the second BAS in which the translation amount used for the frequent kanji code is registered
FIG. 1C shows a basic structure of an E array, and FIG. 1C shows a basic structure of a CHECK array in which frequent kanji codes are registered corresponding to the subscripts of FIG. 1B in a new data structure of the present invention. ing.

As shown in the BASE array structure of FIG. 1A, two kinds of values are put in the BASE array. On the other hand, the value is a conventional parallel movement amount d, which is applied to a kanji code which is not used often (infrequently).
On the other hand, for frequently appearing kanji codes, values different from those (that is, the other of the two values in the BASE array) are registered. This other value corresponds to the second B value in FIG.
This corresponds to a horizontal suffix I1 of the ASE array.

FIG. 7 is a diagram for explaining a 2-byte character code area. FIG. 7A shows a 7-bit code area (Japanese JIS code area, Chinese GB231).
2-80 7-bit code area), and FIG.
Is an 8-bit code area (Japanese EUC code area, Chinese GB2312-80 8-bit code area).

Japanese JIS code and Chinese GB23
As shown in FIG. 7A, the area of the 7-bit code such as the 12-80 7-bit code includes both the first byte (division) and the second byte (point) of the 7-bit code. 33 to 126 addresses (1 to 94). For the same purpose, an 8-bit code region such as a Japanese EUC code region or a Chinese GB2312-80 8-bit code region constitutes an 8-bit code as shown in FIG. 7B. Both the first byte (ku) and the second byte (point) are 161-254
94 addresses (1 to 94).

Therefore, in the present embodiment, the addresses 33 to 126, which are the areas representing the first byte (division) and the second byte (dot) of the 7-bit code, are divided into three addresses.
Classification is made into the categories d1, d2, and d3. In the same manner, the addresses 161 to 254, which are the areas representing the first byte (division) and the second byte (dot) constituting the 8-bit code, are divided into three, and three types of d1,
Classification is performed for d2 and d3.

Therefore, in the present embodiment, when determining the subscript of the second BASE array shown in FIG.
Each has its own translation amount.

That is, as shown in the CHECK array structure of FIG. 1C, even if a character code (that is, a child node) linked to the same character (that is, a parent node) has its code value (7-bit code or 8-bit code) By providing unique translation amounts d1, d2, and d3 in accordance with (a range like a code), it is possible to provide an empty registration location for each kanji code so that they overlap each other on the CHECK arrangement.

In other words, the kanji codes “yen”, “king”, “house”,
In the case of “fire”, “meeting”, “river”, “approximate”, “study”, and “vessel” (all of them are child nodes), depending on their code values, their own translation amounts d1, d2, d3 Provided on CHECK sequence.

Specifically, for the child nodes of the kanji codes "yen", "king", "house", "fire", and "kawa" connected to the parent node "large", these code values are 8K. Since it is included in Ｋ16K, the translation amount = d1 is given on the CHECK array. For the same purpose, for the child nodes of the kanji codes "kai", "approximately", "study", and "vessel" connected to the parent node "large", these code values are included in 16 to 24K. The translation amount = d2 is given on the CHECK array. In this way, the kanji codes "yen", "king", "house", "fire",
"Kai", "river", "about", "study", "vessel" (child node)
Can overlap with each other on the CHECK arrangement to provide an empty registration location for each kanji code.

As a result, compared with the conventional case where the translation amount on the CHECK arrangement is uniformly provided for each kanji code,
The increase in the CHECK arrangement can be suppressed, and the spatial efficiency of the double arrangement can be improved.

As described above, according to the present invention, the key search target is a new data structure which is a further development of the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. Index 1 at the beginning of each character string
A first array having a subscript of 06c, a second array in which information relating to a specific character connected to the beginning of the character string in the character string indicated by the first array is registered, and a key calculated by the translation amount calculating means. The characters of each character string to be searched are set in the first array and the second
A second array in which the index 106c of the beginning of the character string is registered using the sum of the translation amount d required for registration in the array and the value corresponding to the character connected to the end of the character string as a subscript. CHEC by introducing a simple data structure
It becomes possible to give an empty registration location to each kanji code by overlapping each other on the K array, and as a result, without expanding the CHECK array as much as possible, all the kanji codes as keys can be assigned on the CHECK array. It is possible to register in the vacant space at the same time, and it is possible to register in the CHECK array while maintaining the relative positional relationship of each kanji code linked to a certain kanji code without expanding the CHECK array as much as possible. It is possible to avoid as many gaps (sparseness) as possible. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to (First Embodiment) FIG. 2 is a functional block diagram for explaining a first embodiment of the character code registration and search device of the present invention.

The character code registration search device 10 shown in FIG.
Has a function of registering a character string such as a kanji code as a key search target in a double array structure which is a one-dimensional array as a data structure, and searching for the character string.
Frequent character code storage means 103, frequent character code selection means 104, dictionary means 106, category storage means 108, BA
SE arrangement means 102, code classification means 107, parallel movement amount calculation means 109, parallel movement amount storage means 110, key candidate point calculation means 111, second BASE arrangement 105, CHE
The configuration is centered on the CK arrangement means 112.
Each of these components is realized by programming using a microcomputer.

The list means 101 has a function of creating a list of kanji codes frequently used for idioms and outputting a selected character code 103c selected from the kanji code list.

The frequent character code selecting means 104 has a function of outputting a frequency threshold value as to how many kanji codes are to be selected.

The frequent character code storage means 103 has a function of storing the frequent character codes selected from the list means 101 and outputting the selected frequent character codes 103a and the index 106c of the selected frequency character codes.

The dictionary means 106 is a character code dictionary in which idioms composed of kanji codes are registered. The dictionary means 106 divides the work depending on whether or not the character of interest is the beginning of an idiom based on the selected kanji code. Has a function of outputting each class 106a obtained by classifying the kanji codes linked to the frequently appearing kanji codes.

The category storage means 108 has a function of storing the respective categories 106a obtained by classifying the kanji codes linked to the frequently appearing kanji codes from the dictionary means 106.

The BASE arrangement means 102 functions as a second arrangement for calculating the serial number 103b of the selected character code 101c and for storing it in the index 106c of the same kanji code on the BASE arrangement.

The code classifying means 107 has a function of classifying the second kanji code of the idiom with some bits in the kanji code of the second kanji code in order to classify the characters connected to the frequent kanji code at the beginning of the word. have.

The parallel movement amount calculating means 109 calculates an arbitrary parallel movement amount 109a for each class 106a.
It has a function of calculating the minimum translation amount 109a such that any value added to the code value of each kanji code 06a comes to an empty place on the CHECK array.

The parallel movement amount storage means 110 has a function of storing the parallel movement amount 109a from the parallel movement amount calculating means 109 and storing the parallel movement amount 110a in a subscript position corresponding to the index at the beginning of the BASE array means 102. Have.

The key candidate point calculating means 111 calculates, for each class, the parallel movement amount input from the parallel movement amount calculating means,
The sum added to the code value of each kanji code of the same class is CHE
A subscript value is set on the Ck array, an index at the beginning of the word corresponding to the parent of the kanji code of the same class is registered at the subscript position, and the sum value is expressed by ((conventional beginning) + (character of interest)) It has a function as an index of the beginning of the word.

The second BASE array 105 stores the translation amount storage unit 1 based on the code value 107a from the code classification unit 107 and the serial number 103b from the list unit 101.
It functions as a second array for storing the parallel movement amount 110a for each class output by 10.

The CHECK arrangement means 112 stores the parallel movement amount 111a from the parallel movement amount storage means 110 and the internal set value 109a in each kanji code of the class from the parallel movement amount calculation means 109 as an index 106c at the beginning of the CHECK arrangement.
Function as the first array registered in the location of

As described above, according to the character code registration / search apparatus 10 of the first embodiment, a new double array structure which is a one-dimensional array as a conventional high-speed, low-capacity dictionary data structure is further developed. As a simple data structure, the internal setting value 109a in each kanji code of the class from the parallel movement amount calculating means 109 and the parallel movement amount 111a from the parallel movement amount storing means 110 is stored in the CHECK array at the beginning of the index 10.
CHECK array means 112 as a first array to be registered at the position 6c and a second array to be stored in the index 106c of the same kanji code on the BASE array while calculating the additional number 103b of the selected kanji code 101c. BA
SE array means 102, code value 107a from code classification means 107, and serial number 103b from list means 101
Based on the above, a new data structure having a second base array 105 as a second array for storing the parallel movement amount 110a for each class output by the parallel movement amount storage unit 110 is introduced, and Two types of values are registered. One value is applied to a kanji code that is not used so often as a parallel translation amount (low frequency), and the other value is any one of the subscripts of the second BASE array. Applied to the frequent kanji code as a subscript, the subscripts of the second BASE array are divided into three types according to the code values of the characters connected to the frequent kanji code, and each is given its own parallel movement amount, so that they overlap each other on the CHECK array. It is possible to give an empty registration location to each kanji code, and as a result, without expanding the CHECK arrangement as much as possible, It becomes the kanji code can be registered at the same time free on CHECK sequence, also CHECK
CHEC while maintaining the relative positional relationship of each kanji code connected to a certain kanji code without expanding the arrangement as much as possible
It becomes possible to register in the K array, and in addition, it is possible to avoid as many gaps (sparseness) as possible without registering as much as possible. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

FIG. 3 is a flowchart for explaining an embodiment in which a kanji code is registered using the character code registration and search method executed by the character code registration and search device of FIG.

The embodiment of the character code registration and search method shown in FIG. 3 is executed by the character code registration and search device 10 of the first embodiment.
2), frequent character code selection step (step S2), frequent character code storage step (step S2), dictionary step (step S3), classification result storage step (step S9), BA
SE array process (steps S6, S11), code classification process (step S7), translation amount calculation process (step S7)
8, 10), the parallel movement amount storing step (step S8, 1)
0, 11), a key candidate point calculation step (steps S9, 1)
2), second BASE arrangement step (step S9), CHE
It has a logical configuration centering on the CK arrangement process (step S12), and is described in a program code executable by the character code registration and search device 10 of the first embodiment.

The list step (step S2) has a function of creating a list of kanji codes frequently used for idioms and outputting the selected kanji code 103c selected from the kanji code list. This is a processing step mainly executed by the means 101.

Frequent character code selection step (step S2)
Is a processing step which has a function of outputting a frequency threshold value as to how many kanji codes are to be selected, and is executed mainly by the frequently appearing character code storage means 103.

Frequent character code storage step (step S2)
Has a function of storing frequently-used kanji codes selected from the list process (step S2) and outputting a selected frequently-used character code 103a and an index 106c of a selected frequency character code. This is a processing step to be executed.

The dictionary step (step S3) is a character code dictionary in which idioms composed of kanji codes are registered.
It has a function to divide the work depending on whether the character of interest is the beginning of a idiom based on the selected kanji code, and to output each class 106a obtained by classifying the kanji code connected to the frequent kanji code at the beginning of the word, This is a processing step mainly executed by the dictionary unit 106.

The classification result storing step (step S9) has a function of storing each class 106a obtained by classifying the kanji codes linked to the frequently appearing kanji codes at the beginning of the word generated by the dictionary step (step S3). This is a processing step mainly executed by the means 108.

The BASE arrangement step (steps S6 and S11) as the second arrangement step includes the selected character code 1
This is a processing step that has a function of calculating the additional number 103b of 01c and storing the same in the index 106c of the same kanji code on the BASE array, and is executed mainly by the BASE array unit 102.

In the code classification step (step S7), the second kanji code of the idiom is classified by several bits in the kanji code of the second kanji code in order to classify characters connected to the frequent kanji code at the beginning of the word. This is a processing step that has a function of classifying and is executed mainly by the code classifying unit 107.

Parallel movement amount calculating step (steps S8, 1)
0) indicates an arbitrary parallel movement amount 109 for each class 106a.
has the function of calculating the minimum parallel movement amount 109a such that any value obtained by adding a to the code value of each kanji code of the same class 106a comes to an empty place on the CHECK array. This is a processing step executed mainly by the quantity calculation unit 109.

Parallel movement amount storing step (steps S8, S1)
(0, 11) is a parallel movement amount calculating step (steps S8, 1).
0) has the function of storing the generated parallel movement amount 109a and storing the parallel movement amount 110a at a subscript position corresponding to the index at the beginning of the BASE arrangement means 102, with the parallel movement amount storage means 110 as the center. This is the processing step to be executed.

Key candidate point calculation step (steps S9, 1)
2) For each class, the sum of the translation amount input from (Steps S8, 10, 11) and the code value of each kanji code of the same class is used as a subscript value on the CHECk array,
Registers the index of the beginning of the word that is the parent of the kanji code of the same class at the same subscript position, and uses the sum value as the index of the next beginning of ((conventional beginning) + (character of interest)) Then, this is a processing step executed mainly by the key candidate point calculation means 111.

The second BASE arranging step (step S9) as the second arranging step is a code classifying step (step S9).
Based on the code value 107a generated by 7) and the serial number 103b generated by the list process (step S2), the parallel movement amount for each class output by the parallel movement amount storage step (steps S8, 10, 11). 110a has a function of storing
This is a processing step mainly executed by the second BASE array 105.

The CHECK arrangement step (step S12) as the first arrangement step includes a parallel movement amount 111a and a parallel movement amount calculation step (steps S8, 10) generated by the parallel movement amount storing step (steps S8, S11). Has the function of registering the minimum translation amount 109a generated by the CHECK array at the position of the index 106c at the beginning of the CHECK array, and is a processing step mainly executed by the CHECK array unit 112.

As described above, according to the embodiment of the character code registration search method shown in FIG. 3, the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure is further developed. As a new data structure, the parallel movement amount 111a generated by the parallel movement amount storing step (steps S8, 10, 11) and the minimum parallel movement amount 109a generated by the parallel movement amount calculating step (steps S8, 10) are represented by C.
A CHECK arrangement step as a first arrangement step (step S1) to be registered at the index 106c at the beginning of the HECK arrangement
2) and the additional number 103 of the selected kanji code 101c
b, and at the same time, the BAS as a second arraying step is stored in the index 106c of the same kanji code on the BASE array.
A second BASE arrangement step as a second arrangement step for storing the parallel movement amount 110a for each class output by the E arrangement step (steps S6, 11) and the parallel movement amount storage step (steps S8, 10, 11). (Step S9) and a new data structure created by execution are introduced, and two types of values are registered in the BASE array, and one of the values is not used as a parallel translation amount as in the past (infrequently used values). ) Applied to the kanji code, the other value is applied to the frequent kanji code as one of the subscripts of the second BASE array, and the subscript of the second BASE array is divided into three types according to the code values of the characters connected to the frequent kanji code By giving unique translation amounts to each other, it is possible to provide an empty registration location for each kanji code so that they overlap each other on the CHECK array. As a result, all the kanji codes as keys can be registered simultaneously in the empty space on the CHECK array without expanding the CHECK array as much as possible. It is possible to register in the CHECK array while maintaining the relative positional relationship of each kanji code, and in addition, it is possible to avoid as much as possible a gap (sparse) due to failure to register. . Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured,
The storage capacity of a tri-array structure in which a key is added and registered later to expand the memory can be reduced as much as possible.

(Second Embodiment) FIG. 4 is a functional block diagram for explaining a character code registration and search device according to a second embodiment of the present invention.

The character code registration search device 10 shown in FIG.
Has a function of registering a character string such as a kanji code as a key search target in a double array structure which is a one-dimensional array as a data structure, and searching for the character string.
1. Document input means 201, BASE arrangement means 102, registration value judgment means 202, code classification means 107, second BA
The configuration is centered on the SE array 105, the translation amount storage unit 110, the key candidate point calculation unit 111, the CHECK array unit 112, and the key / head prefix matching unit 203. Each of these components is realized by programming using a microcomputer.

The document input means 201 first designates the root of the trie structure as the beginning of a word, sets the end mark # as an end symbol to the beginning of the word W, and then inputs the kanji code b as the character to be searched. And detects the beginning W of the input kanji code b.

The document input means 201 inputs the numerical value 102a from the position corresponding to the initial W or the kanji code index 106c.
Is provided.

The registered value judging means 202 determines whether the numerical value 102a inputted from the BASE arranging means 102 is an additional number of a frequently appearing initial character code, an index of an infrequent initial character code, or a part of the character string. Index 10 of a certain initial W
6c, and when an index exceeding the range of the index constituting the trie is given, this index is output as an additional number 202a of the frequently appearing character code, and at the same time, the numerical value 102a input from the BASE It has a function of outputting the parallel movement amount 202b when the index is not the index 106c of the initial W and the character code is not a frequent character code.

The code classifying means 107 classifies characters connected to the frequent kanji code at the beginning of the word, if the serial number of the initial character code in which the numerical value input from the BASE array means frequently appears is input from the BASE array means, It has a function of classifying by some bits in the kanji code of the kanji code linked to the leading kanji.

The second BASE array 105 includes the additional number 202a of the frequent character code output from the registered value determination means 202.
And has a function of storing the parallel movement amount 105a from a location corresponding to the classification of the code value of the kanji code b.

The parallel movement amount storing means 110 has a function of storing the numerical value 102a as the parallel movement amount when the numerical value 102a input from the BASE arranging means 102 is the index of the initial character code or the index 106c of the initial W which is not frequent. Have.

The key candidate point calculation means 111 checks the CHECK
It has a function of calculating and outputting the sum of the parallel movement amount on the array and the code value of the consecutive character b.

The CHECK arrangement means 112 has a function of inputting a key to a position corresponding to the sum 111a from the key candidate point calculation means 111.

The key / prefix collation means 203 determines whether the key input by the CHECK arrangement means 112 is equal to the index of the initial character code or the index 106c of the initial W, and determines whether the input key is the index of the initial character code. Or index 1 of initial W
It has a function of determining that the idiom is registered in the dictionary when it is determined that it is equal to 06c.

As described above, according to the character code registration / search apparatus 10 of the second embodiment, a new double array structure, which is a one-dimensional array as a conventional high-speed, low-capacity dictionary data structure, is further developed. CHECK arrangement means 112 for inputting a key to a place corresponding to the sum 111a from the key candidate point calculation means 111, and a second value 102a for inputting a numerical value 102a from a place corresponding to the index 106c of the initial W or the kanji code. The BASE array means 102 as an array and the additional number 202a of the frequently appearing character code output from the registered value determination means 202, for storing the parallel movement amount 105a from a position corresponding to the classification of the code value of the kanji code b By introducing a new data structure with the second base array 105 as the second array, the key set can be known in advance. The quasi-static key set like being constitutes a dictionary with a search target, so the storage capacity of the tri-array structured to extend additionally registered later appropriate key can be minimized. As a result, the data is stored in a double array structure (that is, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible, and the tri-array structure can be subjected to high-speed pattern matching using a search key. become.

FIG. 5 is a flowchart for explaining an embodiment in which a kanji code is registered using the character code registration and search method executed by the character code registration and search device of FIG.

The embodiment of the character code registration and search method shown in FIG. 5 is executed by the character code registration and search device 10 of the second embodiment, and includes a document input step (step P).
2, P3), BASE arranging step (step P4), registered value judging step (step P5), code classification step (step P6), second BASE arranging step (step P7), parallel movement amount storing step (steps P7, P8) Key candidate point calculation step (step P9), CHECK arrangement step (step P9), key / prefix verification step (steps P10, p1)
1, P12). Each of these components is realized by programming using a microcomputer.

The document input process (steps P2 and P3)
At first, specify the root of the trie structure as the beginning of the word,
A function of setting an end mark # as an end symbol at the beginning of a word W, subsequently instructing input of a kanji code b as a character to be searched, and detecting the beginning of the kanji code b that has been input, This is a processing step executed mainly by the document input unit 201.

The BASE arranging step (step P4) has a function of inputting a numerical value 102a from a position corresponding to the initial W or the kanji code index 106c, and is a processing step mainly executed by the BASE arranging means 102. .

The registered value determining step (step P5)
BASE arranging step 102 (step P
It is determined whether the numerical value 102a generated in 4) is an additional number of a frequently occurring initial character code, an index of an initial character code that is not so, or an index 106c of an initial W code in the middle of a character string, When an index exceeding the range of the index constituting the trie is given, this index is output as the additional number 202a of the frequently appearing character code, and at the same time, the numerical value 102a generated by the BASE array process 102 (step P4) is This is a processing step that has a function of outputting the parallel movement amount 202b when the index is not the index 106c and is not a frequently appearing character code, and is executed mainly by the registered value determination unit 202.

In the code classification step (step P6), a BA is used to classify characters connected to the frequent kanji code at the beginning of the word.
When an additional number of the initial character code in which the numerical value frequently input from the SE array process 102 (step P4) is input from the BASE array means, classification is performed using several bits in the kanji code of the kanji code linked to the first kanji. Has the function of
This is a processing step mainly executed by the code classification unit 107.

The second BASE arranging step (step P7) as the second arranging step is an additional number 202a of the frequently appearing character code generated by the registered value judging step 202 (step P5), which corresponds to the classification of the code value of the kanji code b. From the location where the parallel movement amount 105a is stored.
This is a processing step mainly executed by the ASE array 105.

Parallel movement amount storing step (steps P7, P
8) is a numerical value 102a when the numerical value 102a generated by the BASE arranging step 102 (step P4) is an index of an infrequent initial character code or an index 106c of an initial W character.
Is a processing step that has a function of storing as a parallel movement amount, and is executed mainly by the parallel movement amount storage unit 110.

The key candidate point calculation step 111 (step P
9) is a character b connected to the amount of translation on the CHECK array
This is a processing step that has a function of calculating and outputting the sum with the code value of the above, and is mainly executed.

The CHECK arrangement step (step P9) as the first arrangement step has a function of inputting a key to a place corresponding to the sum 111a from the key candidate point calculation step 111 (step P9). This is a processing step mainly executed by the means 111.

Key / Initial Verification Step (Steps P10 and p1)
1, P12) determines whether the key generated by the CHECK arrangement step 112 (step P9) as the first arrangement step is equal to the index of the initial character code or the index 106c of the initial W, and A processing step that has a function of determining that an idiom is registered in the dictionary when it is determined that it is equal to the index of the initial code or the index 106c of the initial W, and which is executed mainly by the key / initial verification means 203. is there.

As described above, according to the embodiment of the character code registration search method shown in FIG. 5, the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, is further developed. As a new data structure, the sum 111 from the key candidate point calculation step 111 (step P9)
A CHECK arrangement step 112 (step P9) as a first arrangement step of inputting a key at a position corresponding to a, and a second arrangement step of inputting a numerical value 102a from a position corresponding to an initial letter W or a kanji code index 106c. BASE
Arrangement step 102 (step P4) and registered value determination step 2
02 (step P5) is used as the additional number 202a of the frequent character code generated, and a second translation amount 105a is stored from a location corresponding to the classification of the code value of the kanji code b.
Second BASE arranging step (step P
7), a dictionary is stored in which a quasi-static key set whose key set is known in advance is stored as a search target, and the key is added and registered as needed to expand the dictionary. This makes it possible to minimize the storage capacity of the tri-array structure. As a result, the data is stored in a double array structure (that is, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible, and the tri-array structure can be subjected to high-speed pattern matching using a search key. become. (Third Embodiment) FIG. 6A shows an operation of selecting a specific character executed by the character code registration / search apparatus of FIG. 1 or FIG. 2 based on frequent kanji codes (those having a large number of consecutive kanji codes). FIG. 5 is a diagram for explaining a specific character selecting operation performed by the character code registration and search device of FIG. 1 or FIG. 2 based on the width of a code value of a series of kanji codes. .

The character code registration search method according to the third embodiment is different from the character code registration search method according to the first or second embodiment in that a specific character connected to the beginning of a character string is an idiom. The method is characterized in that a step of selecting a character having a code value width of a continuous kanji code equal to or greater than a predetermined threshold value as a specific character connected to the beginning of a character string is performed.

As a result, more vacant portions that cannot be filled by the conventional method can be filled more.
BASE created in arrangement process 102 (step P4)
An increase in both the array and the CHECK array created in the CHECK array step 112 (step P9) is also moderately suppressed. In addition, since the number of times of processing refers to the second BASE arranging step (step P7), the number of times of processing is increased only once, so that the number of times of processing is almost the same.

As described above, according to the present invention, the size of the original BASE array and the size of the CHECK array are each 64 KW, and the second BASE array is such that the selected kanji code is 0.5 KW and the classification is three. Then 1.5
(= 0.5 × 3) KW. This is only 1/64 of the size of the original BASE and CHECK sequences.
On the other hand, the increase in the size of the BASE array and the CHECK array according to the conventional method is unknown. However, according to the present invention, more vacant portions not filled by the conventional method are filled more, and the increase of both the BASE array and the CHECK array is appropriately suppressed. In addition, since the number of times of processing refers to the second BASE array, the number of times of processing is increased by one, and is substantially the same.

[0126]

According to the first aspect of the present invention, a new data structure is introduced by further developing the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. , Vacant registration points can be given to the respective character codes by overlapping each other on the CHECK array. As a result, all character codes as keys can be replaced with C characters without expanding the CHECK array as much as possible.
You can register in the empty space on the HECK array at the same time.
Without expanding the K array as much as possible, the CHE is maintained while maintaining the relative positional relationship of each character code connected to a certain character code.
It becomes possible to register in the CK sequence, and in addition, it is possible to avoid as much as possible sparse vacancies due to failure to register. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to the invention described in claim 2, according to claim 1
In addition to the effects described in the above, as a new data structure further developed from the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, the translation amount from the translation amount storage means CHECK as a first array for registering an internal set value in each character code of the class from the parallel movement amount calculating means at the index position at the beginning of the CHECK array
Arranging means, BASE array means as a second array for calculating the serial number of the selected character code, and storing in the index of the same character code on the BASE array, code values and a list from the code classifying means A new data structure having a second BASE array as a second array for storing the translation amount of each class output from the translation amount storage means based on the serial number from the means is introduced. Two types of values are registered in the array, one is applied to a character code that is not frequently used as a conventional translation amount, and the other value is one of the subscripts of the second BASE array. Applied to frequent character codes as subscripts,
The subscripts of the ASE array are divided into three types according to the code values of the characters linked to the frequently appearing character codes, and each of them is given its own parallel movement amount, so that the registration positions are vacant for each character code so that they overlap each other on the CHECK array Can be given, and as a result, all character codes as keys can be changed to CH without expanding the CHECK array as much as possible.
It becomes possible to register simultaneously in the empty space on the ECK array, and it is possible to register in the CHECK array while maintaining the relative positional relationship of each character code connected to a certain character code without expanding the CHECK array as much as possible. In addition, it is possible to avoid as much as possible sparse vacancies due to registration failure. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to the third aspect of the present invention, the first aspect
In addition to the effects described in (1), as a new data structure obtained by further developing the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, it corresponds to the sum from the key candidate point calculation means. CHECK to enter the key in the place
Arranging means, BASE arranging means as a second array for inputting a numerical value from a position corresponding to the beginning of a word or an index of a character code, and a character code code By introducing a new data structure having a second base array as a second array for storing the amount of translation from a position corresponding to the value classification, a quasi-static method in which a key set is known in advance is provided. A dictionary in which a key set is stored as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate can be minimized. As a result, the data is stored in a double array structure (that is, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible, and the tri-array structure can be subjected to high-speed pattern matching using a search key. become.

According to the fourth aspect of the present invention, a key search target is obtained as a new data structure obtained by further developing the conventional double array structure which is a one-dimensional array as a high-speed, low-capacity dictionary data structure. A first arranging step in which an index at the beginning of each character string is a subscript, a second arranging step in which information on a specific character connected to the beginning of the character string in the character string indicated in the first arranging step is registered, The translation amount required for registering the character of each character string to be a key search target calculated by the translation amount calculating means in the first arrangement step and the second arrangement step corresponds to the character connected to the end of the character string. By introducing a new data structure formed by using the second arrangement step in which the index of the beginning of the character string is registered using the sum of the value and the index as a subscript,
It becomes possible to give an empty registration place to each character code by overlapping each other on the CHECK array,
As a result, all character codes as keys can be simultaneously registered in a space on the CHECK array without expanding the CHECK array as much as possible, and the relative values of each character code connected to a certain character code can be registered without expanding the CHECK array as much as possible. It becomes possible to register in the CHECK array while maintaining the positional relationship, and in addition, it is possible to avoid as much as possible sparse vacancies due to registration failure. Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured,
The storage capacity of a tri-array structure in which a key is added and registered later to expand the memory can be reduced as much as possible.

According to the invention described in claim 5, according to claim 4,
In addition to the effects described in (1), as a new data structure that is a further development of the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, the prefix of each character string to be searched for a key is A first arrangement step using the index as a subscript,
A second arrangement step in which information relating to a specific character connected to the beginning of the character string in the character string indicated in the first arrangement step is registered; and a key search target character string calculated by the translation amount calculating means. Registers the index of the beginning of the character string using the sum of the amount of translation required for registering the character in the first arrangement step and the second arrangement step and the value corresponding to the character connected to the end of the character string as a subscript A new data structure formed using the second arrangement step described above is introduced, and the registered contents registered in the first arrangement step are added to a specific character code located at the beginning of the idiom, If the registered content registered in the first arrangement step is identified as either a character code or a parallel translation amount of the beginning of a character string, and if the registered content registered in the first arrangement step is an additional number of the first specific character code, the second arrangement step Refer to the sequence number indicated by the serial number in By executing the identification process for obtaining the amount of parallel movement, it becomes possible to give an empty registration place to each character code so as to overlap each other on the CHECK array, and as a result, without expanding the CHECK array as much as possible. , All character codes as keys are CHE
CHECK can be registered in the empty space on the CK array at the same time, and the CHECK array can be maintained without expanding the CHECK array as much as possible while maintaining the relative position of each character code connected to a certain character code
It becomes possible to register in an array, and in addition, it is possible to avoid as much as possible sparse vacancies due to registration failure. Thus, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and a storage capacity of a tri-array structure in which a key is registered and expanded later as appropriate is minimized. CH can be changed for each character code.
Compared to the case of giving the amount of translation on the ECK array,
An increase in the CK arrangement can be suppressed, and the spatial efficiency of the double arrangement can be improved.

According to the inventions set forth in claims 6 to 8, the same effects as the effects set forth in claim 5 can be obtained.

According to the tenth aspect, in addition to the effect of the fourth aspect, the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, is further developed. As a new data structure, the internal setting value of each character code of the parallel movement amount generated by the parallel movement amount storing step and the class generated by the parallel movement amount calculating step is CHEC.
A CHECK arrangement step as a first arrangement step to be registered at the index position at the beginning of the K arrangement, a second number of the selected character code is calculated, and a second number is stored in the index of the same character code on the BASE arrangement. A BASE arranging step as an arranging step and a second BASE as a second arranging step for storing a parallel moving amount for each class output by the parallel moving amount storing step
Introduces a new data structure created by executing the E array process and makes two types of values registered in the BASE array. One of them is a character code that is not frequently used as a parallel translation amount. To the second BAS
Apply the frequently occurring character code as one of the subscripts of the E array, divide the subscript of the second BASE array into three types according to the code value of the character connected to the frequently occurring character code, and give each unique translation amount Thus, it becomes possible to provide an empty registration location for each character code so that they overlap each other on the CHECK array, and as a result, CH
All character codes as keys can be registered in the CHECK array at the same time without expanding the ECK array as much as possible, and the relative values of each character code linked to a certain character code can be registered without expanding the CHECK array as much as possible. It becomes possible to register in the CHECK array while maintaining the positional relationship, and in addition, it is possible to avoid as much as possible sparse vacancies due to registration failure.
Thereby, a dictionary storing a quasi-static key set whose key set is known in advance as a search target is configured, and the storage capacity of a tri-array structure in which keys are registered and expanded later as appropriate is reduced as much as possible. Will be able to

According to the eleventh aspect of the present invention, in addition to the effect of the fourth aspect, the conventional double array structure, which is a one-dimensional array as a high-speed, low-capacity dictionary data structure, is further developed. As a new data structure, a CHECK arrangement step as a first arrangement step of inputting a key to a place corresponding to the sum from the key candidate point calculation step, and a second step of inputting a numerical value from a place corresponding to an index of a head or a character code. A second arraying step for storing a parallel movement amount from a position corresponding to a code value classification of a character code as a serial number of a frequently occurring character code generated by a BASE arraying step as a two-arraying step and a registered value judging step; A dictionary storing a quasi-static key set whose key set is known in advance as a search target by introducing a new data structure having a second BASE arrangement process as Configured, so the storage capacity of the tri-array structured to extend additionally registered later appropriate key can be minimized. As a result, the data is stored in a double-array structure (ie, a tri-array structure), which is a one-dimensional array as a data structure with as small a storage capacity as possible,
This trie array structure can be subjected to high-speed pattern matching using a search key.

According to the twelfth aspect of the present invention, in addition to the effect of the tenth or eleventh aspect, more vacant portions that have not been filled by the conventional method are buried. BASE array to be created and C
The increase in both sequences of the CHECK sequence prepared in the HECK sequence step is also appropriately suppressed. Moreover, the number of processing is 2B
Since the number is increased only once because the ASE arrangement step is referred to, the number of processes is almost the same.

[Brief description of the drawings]

FIG. 1A is a diagram illustrating a character code registration / search apparatus and a character code registration / search method according to the present invention, in which a new data structure is applied to a character code that is not used frequently (infrequently). FIG. 1B shows the basic structure of a BASE array in which the amount of parallel movement and the frequency of occurrence or the serial number are registered. FIG. 1B shows the new data structure of the present invention in which the amount of parallel movement used for frequently appearing character codes is registered. FIG. 1C shows the basic structure of a CHECK array in which frequently appearing character codes are registered corresponding to the subscripts of FIG. 1B in the new data structure of the present invention. Explain.

FIG. 2 is a functional block diagram illustrating a first embodiment of a character code registration and search device according to the present invention.

FIG. 3 is a flowchart for explaining an embodiment in which a character code is registered using a character code registration search method executed by the character code registration search device of FIG. 2;

FIG. 4 is a functional block diagram for explaining a second embodiment of the character code registration and search device of the present invention.

FIG. 5 is a flowchart illustrating an embodiment in which a character code is registered using the character code registration and search method executed by the character code registration and search device of FIG. 4;

FIG. 6 (a) illustrates a specific character selecting operation performed by the character code registration and search apparatus of FIG. 1 or FIG. 2 based on frequently occurring character codes (those having a large number of consecutive character codes). FIG. 3 is a diagram for explaining a specific character selecting operation performed by the character code registration and search device of FIG. 1 or FIG. 2 based on the width of code values of consecutive character codes.

7A and 7B are diagrams for explaining a 2-byte character code area. FIG. 7A shows a 7-bit code area (Japanese JIS code area, Chinese GB2312-80 7-bit code area). FIG. 7B shows an 8-bit code area (Japanese EUC code area, Chinese GB2312-80 8-bit code area).

FIG. 8 is a diagram for explaining the prior art, and FIG.
(A) shows a BASE sequence having a double sequence structure and CHEC
FIG. 8 (b) shows two one-dimensional arrays of K arrays, their BASE array contents and CHECK array contents.
aby #, batchor #, badger #, ba
Trie (tri) in which dge # and jar # are stored
e) A structure (state transition diagram) is shown, and FIG. 8C is a diagram for explaining a parent-child node relationship in a trie structure and a search operation using a double array.

FIG. 9 (a) shows characters “voltage”, “electricity”, “train”, “denno”, “telephone” in the prior art of FIG.
FIG. 9B is a diagram for explaining an operation of additionally registering a idiom composed of a character code connected to the character "den" in the double array structure, and FIG. 9B illustrates an operation of expanding the double array structure. FIG. 9 (c) is a diagram for
It is a figure for explaining operation of enlargement of a trie corresponding to a double arrangement structure of (b).

[0000]

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Character code registration search apparatus 101 ... List means 101b ... Serial number 102 ... BASE array means (second array) 102a ... Numerical value 103 ... Frequently appearing character code storage means 103a ... Frequently appearing character code 103b ... Serial number 103c ... Selected character code 104: Frequent character code selection means 104a ... Frequent character codes 105 ... Second BASE array (second array) 106 ... Dictionary means 106a, 108a ... Classes obtained by classifying character codes linked to frequently occurring character codes at the beginning of words 106b ... Class of frequently occurring initial character codes or character codes linked to the initials of the second and subsequent characters 106c Index of frequency character codes 107 Code classification means 107a Code values 108 Class storage means 109 Parallel movement amount calculation means 109a Parallel movement Amount 110: parallel movement amount storage means 110a: parallel movement amount 11 ... key candidate point calculation means 111a ... minimum value or sum of parallel movement amount 112 ... CHECK arrangement means (first arrangement) 201 ... document input means 202 ... registration value judgment means 202a ... frequently added character code serial number 202b ... parallel movement amount 203: Key / initial verification unit 204: Word unit compression unit

Claims (12)

[Claims] 1. A character code registration / search apparatus that registers a character code string to be searched for a key in a double array structure that is a one-dimensional array as a data structure, and searches for the character string. A translation amount calculating means for calculating a translation amount necessary to register a character in a column; a first array having a prefix of a prefix of each character string to be searched for the key as a subscript; An identification means for identifying a registered value in the first array; a second array in which information on a specific character connected to the beginning of the character string in the character string shown in the first array is registered; Key candidate point calculating means for calculating a sum of a translation amount registered in the first array and the second array and a value corresponding to a character connected to a suffix which is a suffix of the character string; The sum obtained by means as a subscript And a second array in which the index of the beginning of the character string is registered. 2. A list means for preparing a list of character codes frequently used for idioms and outputting a selected character code selected from the list of character codes, and selecting a character code up to what frequency. A frequent character code selecting means for outputting a frequency threshold regarding whether to do, storing a frequent character code selected from the list means,
A frequent character code storage means for outputting an index of the selected frequent character code and the selected frequency character code, and a character code dictionary in which idioms composed of character codes are registered, wherein the character of interest is the selected character A dictionary unit that divides the work according to whether or not the beginning of a idiom based on the code and outputs a class obtained by classifying a character code connected to a frequently occurring character code at the beginning of the word, and a frequent character code of the beginning of the word from the dictionary unit. Class storing means for storing each class obtained by classifying the character codes connected to the character string; and BA for inputting the serial number of the selected character codes
In order to classify the characters connected to the frequently occurring character code at the beginning of the SE array means,
Code classifying means for classifying the second character code of the idiom by some bits in the character code of the second character code; and The value added to the code value of the code is CHECK
A translation amount calculating means for calculating a minimum translation amount so as to come to a vacant position on the array; storing the translation amount from the translation amount calculating means; Means for storing a translation amount stored in a subscript position corresponding to the index at the beginning of the word, and for each of the classes, the translation amount input from the translation amount calculating means is added to the code value of each character code of the class The sum thus obtained is used as a subscript value on the CHECk array, and the index of the head of the character which is the parent of the character code of the same class is registered at the position of the subscript, and the value of the sum is ((conventional head) + (character of interest))
Key candidate point calculation means as an index of the next word prefix, and each of the parallel movement amount storage means based on the code value from the code classification means and the serial number from the list means. A second BASE array as the second array for storing the translation amount for each class; and the minimum translation amount from the translation amount calculation means as each of the classes from the translation amount calculation means. 2. The character code registration and search device according to claim 1, further comprising: a CHECK arrangement means as the first arrangement for registering an internal set value in a character code at a position of an index at the beginning of the CHECK arrangement. 3. At the beginning, the root of the trie structure is specified as the beginning of a word, and an end mark # as an end symbol is set at the beginning of the word. Subsequently, the input of a character code as a character to be searched is instructed. Document input means for detecting the beginning of a given character code, BASE array means for inputting a numerical value from a position corresponding to the beginning or an index of the character code, and a numerical prefix input from the BASE array means, It is determined whether it is an additional number of the character code, an index of the initial character code that is not the other, or an index of the first letter in the middle of the character string, and an index exceeding the range of the index constituting the trie is given. In this case, this index is output as an additional number of the frequent character code, and when the numerical value input from the BASE arrangement means is not the index at the beginning of the word and not the frequent character code. A registration value determination means for outputting a parallel movement amount in order to classify a character leading to prefix frequently appearing character codes,
Code classification means for classifying with some bits in the character code of the character code connected to the leading kanji when the serial number of the initial character code in which the numerical value inputted from the BASE arrangement means frequently appears is inputted from the BASE arrangement means A second BASE array for storing the parallel movement amount from a location corresponding to the classification of the code value of the character code as an additional number of the frequent character code output from the registered value determination means; When the input numerical value is an index of an infrequent initial character code or an index of an initial word, a translation amount storing means for storing the numerical value as a translation amount, and for each of the classes, an input from the translation amount calculating means. The sum of the translation amount and the code value of each character code of the same class is used as a subscript value on the CHECk array, and the parent of the character code of the same class is placed at the same subscript position. Register an index of the prefix striking, the value of the sum ((conventional prefix) + (character interest))
Key candidate point calculation means as an index of the next word prefix consisting of: a CHECK arrangement means for inputting a key to a place corresponding to the sum from the key candidate point calculation means; and a key input by the CHECK arrangement means, Key / head matching means for judging whether or not the input key is equal to the initial code or initial index, and judging that the idiom is registered in the dictionary when determining that the input key is equal to the initial code or initial index; The character code registration search device according to claim 1, comprising: 4. A method of registering and searching for a character code character string to be searched for a key in a double array structure, which is a one-dimensional array as a data structure, and searching for a character code. A translation amount calculating step of calculating a translation amount necessary for registering a character string; a first arraying step in which an index at the beginning of each character string to be searched for a key is used as a subscript; An identification step of identifying a registration value; and a second step of registering information relating to a specific character connected to the beginning of the character string in the character string indicated in the first arrangement step.
An arrangement step; a key candidate point calculation step of calculating a sum of a translation amount registered in the first arrangement step and the second arrangement step and a value corresponding to a character connected to the end of the character string; A second arrangement step of registering an index at the beginning of the character string using the sum obtained in the key candidate point calculation step as a subscript, and a second arrangement step. 5. The method according to claim 1, wherein the identifying step includes registering the registered contents in the first arrangement step.
A step of identifying one of the serial number I of the specific character code located at the beginning of the idiom, the other leading character code, and the amount of parallel translation of the beginning of the character string; When the registered content is identified as the serial number I of the head specific character code, a step of obtaining a translation amount by referring to an array location indicated by the serial number I in the second arraying step. The character code registration search method according to claim 4, characterized in that: 6. The method according to claim 1, further comprising a step of referring to the second arrangement step based on the additional number I of the leading specific character code of the character string and a classification of a character code connected to the leading specific character code. Item 5. The character code registration search method according to Item 5. 7. The character code registration according to claim 6, wherein the step of classifying a character code linked to the leading specific character code in the second arrangement step uses a code value of the character code. Search method. 8. The method according to claim 4, wherein the second arranging step includes a step of selecting a character frequently used in forming a idiom as a specific character connected to the beginning of the character string. Character code registration search method. 9. In the second arrangement step, when the same character is an idiom, a character having a code value width of a continuous character code equal to or greater than a predetermined threshold is selected as a specific character continuous to the beginning of the character string. 5. The method according to claim 4, further comprising the steps of: 10. A list process for creating a list of character codes frequently used in idioms and outputting a selected character code 103c selected from the list of character codes; A frequent character code selecting step of outputting a frequency threshold regarding whether to select, and storing the frequent character code selected from the list step,
A frequent character code storage step of outputting an index of the selected frequent character code and the selected frequency character code, and a character code dictionary in which idioms composed of character codes are registered, wherein the character of interest is the selected character A dictionary step of dividing each work based on whether or not the beginning of a idiom based on the code and classifying a character code connected to a frequently occurring character code at the beginning of the word, and outputting a class obtained by the dictionary step; A classification result storing step of storing each class obtained by classifying the character codes connected to the codes; calculating the serial number of the selected character codes; and storing the serial numbers in the index of the same character codes in the BASE array A BASE arrangement step as the second arrangement step; and
A code classification step of classifying the second character code of the idiom by some bits in the character code of the second character code; and an arbitrary parallel movement amount 109a for each of the classes 106a.
A translation amount calculating step of calculating a minimum translation amount 109a such that any value obtained by adding to the code value of each character code of the same class 106a comes to an empty place on the CHECK array, A parallel movement amount storing step of storing the parallel movement amount generated by the parallel movement amount calculating step, and storing the parallel movement amount at a subscript position corresponding to the index of the beginning of the BASE arrangement means; The sum of the parallel movement amount input from the parallel movement amount calculation means and the code value of each character code of the same class is used as a subscript value on the CHECk array, and the prefix corresponding to the parent of the character code of the same class at the subscript position Is registered, and the value of the sum is ((conventional prefix) + (character of interest))
A key candidate point calculation step as an index of the next prefix consisting of: the parallel movement amount storage step outputs based on the code value generated by the code classification step and the serial number generated by the list step. A second BASE arrangement step as the second arrangement step for storing the translation amount for each class to be performed; and a translation amount and a translation amount calculation step generated by the translation amount storage step. CHECK as the first arrangement step of registering an internal set value in each character code of the class at the index position at the beginning of the CHECK array
5. The character code registration search method according to claim 4, further comprising a K arrangement step. 11. At the beginning, the root of the trie structure is specified as the beginning of a word, and an end mark # as an end symbol is set at the beginning of the word. Subsequently, an input of a character code as a character to be searched is instructed. A document input step of detecting the beginning of a given character code, a BASE arrangement step of inputting a numerical value from a position corresponding to the beginning or an index of the character code, and a numerical value generated by the BASE arrangement step as the second arrangement step Is the additional number of the initial character code that frequently appears, is it the index of the initial character code that is not, or is it the index of the initial character in the middle of the character string, and determines the range of the index that constitutes the trie When an index exceeding the index is given, this index is output as an additional number of the frequently appearing character code, and the numerical value generated by the BASE arrangement process is not the index at the beginning of the word and frequently appears. A registration value determination step of outputting a parallel movement amount if not character codes, in order to classify a character leading to prefix frequently appearing character codes,
When the additional number of the initial character code in which the numerical value input from the BASE arrangement step frequently appears is input from the BASE arrangement means,
A code classification step of classifying with some bits in the character code of the character code connected to the leading kanji, and an additional number of the frequent character code generated by the registered value determination step, which corresponds to the classification of the code value of the character code. From the point
A second BASE arranging step as the second arranging step for storing the amount of parallel movement; and, if the numerical value generated by the BASE arranging step is an index of an infrequent initial character code or an index of an initial word, the numerical value is parallelized. A parallel movement amount storing step of storing the movement amount as a movement amount, and adding, for each of the classes, the parallel movement amount input from the parallel movement amount calculating means to the code value of each character code of the class, a subscript on the CHECk array. And the index of the beginning of the letter corresponding to the parent of the character code of the same class is registered in the subscript position, and the value of the sum is ((conventional beginning) + (character of interest))
A key candidate point calculation step as an index of the next initial word consisting of: a CHECK arrangement step as the first arrangement step of inputting a key to a place corresponding to the sum from the key candidate point calculation step; and the first arrangement CHECK arrangement as a process It is determined whether the key generated by the process is equal to the initial character code or the index of the initial. If it is determined that the input key is equal to the initial code or the index of the initial, the idiom is added to the dictionary. 5. The character code registration search method according to claim 4, further comprising a key / prefix matching step of determining that the character code has been registered. 12. When a specific character connected to the beginning of a character string is an idiom, a character whose code value width is equal to or larger than a predetermined threshold is specified as a specific character connected to the beginning of the character string. The method according to claim 10 or 11, further comprising a step of selecting.