JP5237400B2 - Search device - Google Patents

Description

  The present invention relates to a search device. In particular, the present invention relates to a search device that searches a given character string for a set of a plurality of registered character strings and detects whether the given character string is a registered character string.

  A technique for detecting whether a character string is a registered character string is a basic technique in information processing and is used in a wide technical field. For example, in banking business, it is used to confirm whether the addressee of a transfer from another bank, particularly a foreign bank, has an account in the bank.

  As one of methods for detecting whether a character string is a registered character string, a technique using a B-tree is known (see, for example, Patent Document 1). In the B-tree, as shown in FIG. 10, a pointer to a child node and a value are stored in nodes constituting the tree structure, and a value smaller than “23” (for example, before “23” in dictionary order). Is stored in slot 1001, and a pointer to a child node storing a value larger than “23” and smaller than “71” is stored in slot 1002, and is larger than “71”. A pointer to the child node storing the value is stored in the slot 1003. When a certain value is given, comparison is made in order with “23” and “71”, and if it is equal to “23” and “71”, it is determined that the registered value is obtained, and “23” and “71” Otherwise, the child node is moved to the child node according to the comparison result.

  Further, when determining whether or not the given character string is included in the character string table, the character string is extracted from the character string table, the first character of the given character string and the extracted character string are compared, If the lengths are the same, the lengths are compared. If the lengths are the same, a method of comparing characters from the end of the character string is known (for example, see Patent Document 2).

  Also, a technique is known in which the length of a word is recorded together with the word to eliminate duplication of the word when performing a full-text search of a document that is a kind of character string (see, for example, Patent Document 3).

JP-A-8-87511 Japanese Patent Laid-Open No. 7-85070 JP 2000-200287 A

  In the conventional technique using a B-tree or the like, search efficiency is improved by narrowing down registered character strings. However, it is necessary to sequentially compare character strings within the narrowed range. For this reason, in order to confirm that no character string is registered, all character strings within the narrowed range are compared. Therefore, although it can be detected with little effort to detect that the character string is registered, it takes time to detect that the character string is not registered.

  As one of the objects of the present invention, there is provided a search device and the like for performing detection of whether a given character string is registered and detection of not being registered with little effort.

  Therefore, as one embodiment of the present invention, there is tree-structured data configured by converting each of a plurality of character strings into a sequence of one or a plurality of tokens, and merging a common sequence of tokens from a start end. An index memory that stores tree structure data in which maximum and minimum lengths to the end nodes are associated with the nodes of the structure data, and a tokenizer that converts a character string to be searched into a sequence of one or more tokens The search is performed by selecting a child node of the node of the tree structure data stored in the index memory while scanning the token row converted by the tokenizer and searching for a route starting from the root node. A search unit that detects whether or not a character string exists in the plurality of character strings, wherein the search unit is a child node of a node. If there is no child node including the number of unscanned tokens within the range of the maximum value and the minimum value of the length to the end node, the character string to be searched does not exist in the plurality of character strings. Provided is a search device for detecting

  More specifically, the search unit includes a token string holding unit, a token number holding unit, a token number holding unit, and a node information holding unit, and the token string holding unit is a sequence of tokens converted by the tokenizer. The token number holding unit holds the number of unscanned tokens, the token number holding unit holds the number of the token being scanned in the token sequence, and the node information holding unit Number of nodes that hold information about the node that has reached, and that are held by the token number holding unit within the range of maximum and minimum lengths to the end node among the child nodes of the currently reached node If there is no child node including, it is detected that the character string to be searched does not exist in the plurality of character strings.

  Further, according to an embodiment of the present invention, there is tree-structured data configured by converting each of a plurality of character strings into a sequence of one or a plurality of tokens and merging a common sequence of tokens from a start end, In the structure data node, the tree structure data in which the maximum value and the minimum value of the length to the end node are associated is stored in the storage unit, and the character string to be searched is converted into one or a plurality of token strings. The character to be searched is searched by selecting a child node of the node of the tree structure data stored in the index memory while scanning the converted token sequence and searching for a route starting from the root node. A program for detecting whether a sequence exists in the plurality of character strings, and a maximum value and a minimum value of a length from a child node of a node to a terminal node If the number of unscanned token exists whatever child nodes within the range, to provide a program for detecting that the string to the search does not exist in said plurality of strings.

  Further, according to an embodiment of the present invention, there is tree-structured data configured by converting each of a plurality of character strings into a sequence of one or a plurality of tokens and merging a common sequence of tokens from a start end, The structure data node is a computer operation method in which the tree structure data in which the maximum value and the minimum value of the length to the end node are associated with each other is stored in the storage unit. Convert to a sequence of one or more tokens, scan the converted sequence of tokens, select a child node of the node of the tree structure data stored in the index memory, and search for a route starting from the root node To detect whether the character string to be searched exists in the plurality of character strings, and the maximum value and the minimum value of the length from the child node of the node to the terminal node If the number of unscanned token exists whatever child nodes within the range, including the character string to be the search does not exist in said plurality of strings, provides a method of operating a computer.

  According to the present invention, it is detected that the given character string is not registered at an early stage of searching whether the given character string is a registered character string, and the given character string is registered. It is possible to detect the presence and the absence of registration with less effort than before.

The functional block diagram of the search device which concerns on one Embodiment of this invention. FIG. 6 is an example of a data structure corresponding to a token column in which a character string is converted. An example of the data structure for comprising the tree structure data concerning one embodiment of the present invention. An example figure of a tree structure corresponding to a plurality of character strings and a plurality of character strings concerning one embodiment of the present invention. The flowchart of the process which comprises the tree structure data which concerns on one Embodiment of this invention. The functional block diagram of the search part of the search device which concerns on one Embodiment of this invention. The flowchart of the process of the search part of the search device which concerns on one Embodiment of this invention. An example figure of the tree structure concerning one embodiment of the present invention, and an example figure of processing of a search part. An example figure of tree structure data concerning one embodiment of the present invention, and a data structure which realizes it. An example figure of B-tree.

  Hereinafter, embodiments for carrying out the present invention will be described as embodiments. Note that the present invention is not limited to the following embodiments, and can be implemented with various modifications.

(Embodiment 1)
FIG. 1 is a functional block diagram of a search device according to an embodiment of the present invention. The search device 100 includes an index memory 101, a tokenizer 102, and a search unit 103.

(Index memory)
The index memory 101 stores one or a plurality of tree structure data. The tree structure represented by one or more tree structure data stored in the index memory 101 satisfies the following two conditions.
(Condition 1) Each of a plurality of character strings is converted into one or a plurality of token strings, and a common token string is merged from the start end.
(Condition 2) The maximum value and the minimum value of the length to the end node are associated with the nodes of the tree structure.

(Definition regarding tree structure)
In this embodiment, the tree structure is a mathematically defined graph structure, and the graph structure has (a) one or a plurality of nodes, and (b) an edge having a start point and an end point as nodes. The starting point node and the ending point node are connected by (c), (c) is simply connected, and (d) there is no edge constituting the loop. Data representing a tree structure is defined as tree structure data. Tree structure data can be represented by a data structure representing nodes and edges. An example of such a data structure is a data structure described later with reference to FIG.

Inductively, (1) while defining a tree structure, simultaneously (2) a root node, (3) a terminal node, and (4) a maximum value and a minimum value of the length from the root node to the terminal node It becomes as follows.
(Definition 1) (1) One node has a tree structure. (2, 3) The node is a root node of the tree structure and a terminal node. (4) In this tree structure, the maximum value and the minimum value of the length from the root node to the terminal node are both 1.
(Definition 2) (1) one or more of the tree _T 1, ..., if there is _{T n,} the tree structure _T 1, ..., a new node not one of the nodes of _{T n} p and the tree structure _{T 1,} ... , T _n are connected to each root node by n edges starting from the node p and ending at each root node, which is a new tree structure. (2) The root node of this new tree structure is node p, (3) The end nodes of this new tree structure are the respective end nodes of tree structures T ₁ ,..., T _n , (4) maximum value from the root node to the terminal node length of the new tree structure, {the maximum length of the root node of T ₁ to the end node, ..., the length from the root node of T _n to the end node a value obtained by adding 1 to the maximum value of the maximum value}, the minimum value of the length from the root node of the new tree to the end node, the length of the to the end node from the root node of {T ₁ minimum value of, ..., is a value obtained by adding 1 to the minimum value of the minimum value of the length from the root node of T _n to the end node}.

From any node q of the tree structure, there is an edge e ₁ to the end point node r ₁ to the start point node q, there is an edge e ₂ of the end point node r ₂ to the start point node r _1, ..., node r _(n-1) when there is an edge en to the end point node _{r n} a start point of the column _{q, r} 1, r 2 _{nodes, ..., r (n-1),} and starting a "node q to _{r n} It defines a path "to the length from node q to node r _n is defined as" n ".

(Definition for token columns)
FIG. 2 is a diagram for explaining conversion of a character string into a sequence of one or more tokens. A token refers to a data structure representing a word that is a character or a sequence of characters. Converting a character string into a string of one or more tokens means converting the character string into a sequence of data structures representing the characters or words that are its constituent elements. Converting a character string into a token string is sometimes referred to as lexical analysis.

  As an example, FIG. 2 illustrates that when the character string “TY HK BMB NKYM” is converted into a token string, the characters “”, “HK”, It shows that the four tokens corresponding to “BMB” and “NKYM” are delimited.

  Each of (a) and (b) of FIG. 2 shows an example of a data structure representing a sequence of tokens when a character string to be converted is converted into a token string. In FIG. 2A, an array including five elements (slots) is prepared. In the slot 201, a pointer to the character string “TY” (the address of the memory in which “TY” is stored), and in the slot 202. Stores a pointer to the character string “HK”, a slot 203 stores a pointer to the character string “BMB”, a slot 204 stores a pointer to the character string “NKYM”, and a slot 205 stores a token string. Indicates a state in which a value indicating the end of (for example, NULL) is stored. Instead of an array of five slots, a data structure in which five lists are connected in series may be used.

  FIG. 2B shows that an array of five slots is prepared, as in FIG. 2A, but each slot stores a character string “TY HK BMB NKYM”. The address in the memory area 210 and the length of the character string represented by the token are stored. That is, the slot 211 stores an address where “T”, which is the start character of “TY” in the area 210, and 2 which is the number of characters “TY”, and the slot 212 stores the address of the area 210. An address storing “H”, which is the start character of “HK”, and “2”, which is the number of characters of “HK”, are stored. In slot 213, the start character which is the start character of “BMB” in area 210 The address where the character “B” is stored and the number 3 of the characters “BMB” are stored, and the slot 214 stores “N” which is the start character of “NKYM” in the area 210. And 4 which is the number of characters “NKYM” are stored. The slot 215 stores a value that is the end of the token string and, for example, 0.

  When converting a character string into a token string, as shown in FIG. 2, the character string can be delimited by a character representing a space or another symbol, and changed into a token string as a word. In addition, when a character representing a space or other symbol is not included, or when a character representing a space or other symbol is not used as a delimiter, a predetermined word is added to the beginning of the character string (the first character of the character string). ) To obtain the token data structure. For example, when the predetermined words are “TY” and “HK”, the character string “HKTY” can be converted into a token string corresponding to “HK” and “TY”. In addition, the character string “TYKTJMHK” is not a word for which “KTJM” is determined in advance, but “TY” and “HK” before and after are predetermined words. , “TY”, “KTJM”, and “HK”.

  It is also possible to convert each character constituting the character string into a token string in association with one token. For example, the character string “TYHK” can be converted into a token string of characters “T”, “Y”, “H”, and “K”. In general, depending on the character code, not all of one character is composed of the same number of bytes. Therefore, when using a character code in which all one character is not composed of the same number of bytes, the character string can be converted into a token string by using one character as one token.

  In the token sequence, the first token is referred to as a “start end of the token sequence”. For example, in the example of FIG. 2, the token corresponding to “TY” is the start end of the token string.

(Definition regarding the tree structure formed by merging token sequences)
In the above (Condition 1), “merging a common token sequence from the start end” refers to the following processing. If a column _{A 1 A 2 ... A n C} 1 C 2 ... C k of the first row of the token _{A 1 A 2 ... A n B} 1 B 2 ... B m and the second token is given, this 2 In the sequence of two tokens, a partial sequence A ₁ A ₂ ... An of _n tokens from the start end A ₁ is common. Therefore, the node corresponding from A ₁ to each token of A _n sequentially connected by an edge in series, a node corresponding to A ₁ and a root node, creating a tree structure to the node corresponding to A _n-terminal node . Then, the child nodes of the node corresponding to A _n, and nodes corresponding to the nodes and C ₁ corresponding to B _1, in turn connected by edges in series node corresponding from B ₁ to B _m each token, Further, it means that nodes corresponding to each of C ₁ to C _k are connected in series with edges.

  Note that, in the two token sequences, when there is no common token sequence from the start end, two different tree structures are configured. For example, when D and G, which are tokens at the start end, are different as in DEF and GEF, a tree structure in which nodes corresponding to token D, token E, and token F are connected in series with edges in order, A tree structure in which nodes corresponding to the token G, the token E, and the token F are connected in series with edges in order is configured.

  If a null character string with 0 characters is present at the beginning of the character string and the leading null character string is a token, a single tree structure is generated even when there is no common token string from the start end. Even in such a case, one embodiment of the present invention is applicable.

(Definition regarding subtree)
In the following description, regarding the token sequence D ₁ D ₂ ... D _n , a tree structure obtained by sequentially connecting the nodes corresponding to D ₁ to D _n in series with edges is referred to as the token sequence D ₁ D ₂ . A tree structure corresponding to D _n is assumed. Further, a path obtained by tracing the edge node in order from the start point of the edge with the root node of the tree structure as a start point is referred to as a subtree of the tree structure. For example, as a sub-tree of a tree structure corresponding to the token sequence E ₁ E ₂ E ₃ E ₄ E _5, it corresponds to E ₁ , E ₁ E ₂ , E ₁ E ₂ E ₃ and E ₁ E ₂ E ₃ E ₄ A tree structure that can be cited.

  In addition, when there is already a tree structure formed by merging a common token sequence from the start end, adding a new token sequence is a part from the start end of the new token sequence. If the tree structure corresponding to the column is a subtree of one of the existing tree structures, the root node of the tree structure corresponding to the token column other than the subsequence is the last node of the subtree. It is a child node of. If the tree structure corresponding to the start side portion of the new token sequence is not any subtree of the existing tree structure, the tree structure corresponding to the new token sequence is added.

(Example of data structure representing nodes and edges)
In the present embodiment, as described in (Condition 2), the maximum value and the minimum value of the length to the end node of the tree structure with the node as the root node are associated with each node constituting the tree structure. It has been. Therefore, the tree structure data stored in the index memory 101 is stored, for example, representing a node and an edge starting from the node by the data structure shown in FIG.

  In FIG. 3, one data structure includes four slots 301 to 304, and the first slot 301 stores characters or character strings represented by tokens. The second slot stores the maximum value of the length to the end node of the tree structure with the node as the root node. The third slot stores the minimum value of the length to the end node of the tree structure with the node as the root node. The fourth slot 304 stores child node information, thereby representing an edge. The first slot 301 does not need to store the character or character string itself represented by the token, but has information for representing the character or character string represented by the token, such as a pointer to the character or character string represented by the token. It only has to be stored.

  The child node information stored in the slot 304 is information necessary for referring to the data structure representing the child node. For example, an address of a memory in which a data structure corresponding to a child node is stored. If there are a plurality of child nodes of one node, a plurality of memory addresses are stored in the fourth slot 304. Further, in the data structure representing the end node, the information of the child node is not stored.

(Example of a tree structure corresponding to a token sequence)
FIG. 4 shows an example of a plurality of token sequences and a tree structure formed by merging common token sequences from the plurality of token sequences from the start end.

  As shown in FIG. 4A, ABCD is used as the first token sequence, ABDEC is used as the second token sequence, ACEBEF is used as the third token sequence, and ACEBGH is used as the fourth token sequence. Suppose you are given. A tree structure formed by merging a common token sequence from these token sequences from the start end, as shown in FIG. 4B, uses data structures 401-414 corresponding to each node. Composed. For example, the data structure 401 corresponds to the node of token A. Since the token A appears at all the start ends of a plurality of token strings, the maximum value of the length to the end node of the node corresponding to the data structure 401 is 6, which is the number of ACEBEF and ACEBGH tokens, and is the minimum value. Is 4 which is the number of tokens of ABCD.

  As child nodes of the node corresponding to the data structure 401, there are a node corresponding to the data structure 402 and a node corresponding to the data structure 408. The former corresponds to token B, and the latter corresponds to token C. For this reason, as the value of the fourth slot of the data structure 401, an address in which each of the data structure 402 and the data structure 408 is stored and a value NULL indicating the end of the column of addresses are stored. Corresponding to BDEC and BCD, 4 and 3 are stored in the data structure 402 as maximum and minimum values up to the end node of the tree structure having the node corresponding to the data structure 402 as a root node. Are stored as a data structure 403 corresponding to the root node of the tree structure corresponding to the token string CD and a data structure 405 corresponding to the root node of the tree structure corresponding to the node string DEC. Is stored. Hereinafter, the data structures 404-414 are described in the same manner.

  In particular, in the data structures 404, 407, 412, and 414 corresponding to the end node, the maximum value and the minimum value of the length to the end node of the tree structure having the corresponding node as the root node are the same, and 1 is there. Therefore, whether or not the node is the terminal node can be determined based on whether or not the maximum value and the minimum value of the length to the terminal node are both 1.

(An example of a flowchart of processing for merging common token sequences from the start)
The method for constructing a tree structure configured by using a node corresponding to a token by merging a common token sequence from the start end is generally described, but an example specifically showing it in a flowchart is as follows. FIG.

  As the processing in step S501, the variables L and M are initialized. That is, the number of tokens in the sequence of tokens to be merged is substituted for variable L, and 1 is substituted for variable M. In the flowchart of FIG. 5, the token string is scanned, the token character strings are acquired in order from the start end, and compared with the node character strings in order from the root node of the tree structure. At this time, the variable L represents the number of unscanned tokens, and the variable M represents what number of tokens are currently focused on from the start end.

  As processing in step S502, the array N is initialized. The array N stores path information starting from the root node of the tree structure. Therefore, in step S502 in which M is 1, the Mth element N [M] of the array N has a data structure corresponding to the root node of the tree structure, and the character string of the token at the start end is set to the first character string. Substitute the address of the data structure stored in the slot. Note that if there is no data structure corresponding to the root node of such a tree structure, NULL is substituted for N [M].

  In step S503, it is determined whether the value of N [M] is NULL. If the value of N [M] is NULL, there is no common token sequence from the start end. Therefore, the process proceeds to step S504, and a tree structure corresponding to the token string is configured and added.

  In step S503, if the value of N [M] is not NULL, a common token string is set from the start end. Therefore, the process proceeds to step S505, and the maximum value and the minimum value of the node represented by N [M] are adjusted. That is, if the value of L exceeds the maximum value of the length to the end node of the node represented by N [M], L is stored as the maximum value of the node represented by N [M]. If the value of L is less than the minimum value of the length of the node represented by N [M] to the end node, L is stored as the minimum value of the node represented by N [M].

  Since one token has been processed as the processing of step S506, the value of L is decreased by 1, and the value of M is increased by 1 in order to process the next token. In step S507, it is determined whether or not the value of L is equal to zero. If the value of L is equal to 0, nodes corresponding to all tokens in the token sequence exist in the tree structure, and the process ends. If the value of L is not equal to 0, the process proceeds to step S508.

  In step S508, the address of the node having the character string of the Mth token among the child nodes of the node represented by N [M-1] is substituted into the variable N [M].

  In step S509, it is determined whether the value of N [M] is NULL. If the value of N [M] is NULL, the process proceeds to step S510, a tree structure corresponding to the Mth and subsequent token strings is generated, and the root node of N [M−1] Add as a child node.

  If the value of N [M] is not NULL in the process of step S509, the process returns to step S505 to process the Mth token and the node corresponding to N [M].

  When a plurality of character strings are given, the index memory 101 stores tree structure data obtained by executing, for example, the flowchart of FIG. 5 for each character string.

  When a plurality of character strings are given to the index memory 101 and tree structure data corresponding to the tree structure satisfying the conditions 1 and 2 is stored, a new character string exists in the plurality of character strings. Whether or not to do so can be detected as follows. That is, when a new character string is converted into a token string and the flowchart of FIG. 5 is executed, whether or not N [M] is a terminal node when the value of the variable L becomes 0 in step S507. It can be detected depending on why. Therefore, in the following, when a plurality of character strings are given and tree structure data corresponding to the tree structure satisfying conditions 1 and 2 is stored, a new character string is included in the plurality of character strings. The process for detecting whether or not it exists will be described in more detail. The new character string is referred to as “character string to be searched”.

(Tokenizer)
The tokenizer 102 converts a character string to be searched into a string of one or more tokens. The converted token sequence is stored in the memory of the search device 100. For example, the address of the stored memory is transmitted to the search unit 103.

(Search part)
The search unit 103 scans the token string converted from the character string searched by the tokenizer 102, selects a child node of the node of the tree structure data stored in the index memory 101, and starts from the root node Search for. That is, the search unit 103 sequentially scans the token string converted from the character string searched by the tokenizer 102 from the start end, and acquires information on the token being scanned. Then, using the acquired token information, a node is selected from the child nodes of the node being scanned in the tree structure represented by the tree structure data stored in the index memory 101. This is called “selecting a child node”.

  When selecting a child node, a selection is made so that the number of unscanned tokens is included in a range that is not more than the maximum value and not less than the minimum value of the length to the end node of the selected child node. If the number of unscanned tokens exceeds the maximum length of any child node to the end node, or falls below the minimum length of any child node to the end node, Since the number of tokens in the character string to be searched is not the same as the number of tokens in any character string in the plurality of character strings, it can be immediately detected that the character string to be searched is different from any character string in the plurality of character strings.

  Therefore, the search unit 103 may include a token string holding unit 602, a token number holding unit 603, a token number holding unit 604, and a node information holding unit 605 as shown in FIG. The token string holding unit 602 holds a token string obtained by converting a character string to be searched by the tokenizer. For example, the addresses of the arrays shown in FIGS. 2A and 2B are stored using a register or the like. The token number holding unit 603 holds the number of tokens in the token string obtained by converting the character string to be searched by the tokenizer. For example, the number of tokens is stored using a register or the like. The token number holding unit 604 holds the number of the token currently being scanned in the token string. For example, the token number is stored using a register or the like. The node information holding unit 605 stores information related to the data structure corresponding to the currently scanned node of the tree structure data. For example, the address of the data structure corresponding to the node is stored using a register or the like.

  FIG. 7 is a flowchart for explaining the processing flow of the search unit 103. The variable L is a variable that holds the number of unscanned tokens. The variable M is a variable indicating what number token is being scanned. Therefore, in step S701, the number of tokens in the token string transmitted from the tokenizer is substituted into the variable L. Also, 1 is assigned to the variable M. The variable L corresponds to the token number holding unit 603, and the variable M corresponds to the token number holding unit 604.

  As a process of step S702, information on the root node that has the character string of the first token, the value of the variable L, and the range between the maximum value and the minimum value of the length to the end node is stored in the variable N. substitute. The variable N corresponds to the node information holding unit 605.

  In step S703, it is determined whether the value of N is NULL. If the value of N is NULL, there is no root node that has the character string of the first token and is within the range between the maximum value and the minimum value of the length to the end node. The character string to be registered is not registered. That is, there is no character string to be searched for among a plurality of character strings. Therefore, the process proceeds to step S704, and it is determined that no character string is registered.

  If the value of N is not NULL in step S703, the process proceeds to step S705, and one token is scanned. Therefore, the value of the variable L that stores the number of unscanned tokens is decreased by 1, and the next token is In order to scan, the value of the variable M is increased by one.

  In step S706, it is determined whether the value of the variable L is equal to zero. If the value of the variable L is equal to 0, all tokens in the token row have been scanned, and the process proceeds to step S707. In step S707, it is determined whether the variable N represents a terminal node. If N is not a terminal node, the process proceeds to step S708 to determine that no character string is registered. If it is determined in step S707 that N is a terminal node, the process proceeds to step S709 to determine that the character string is registered.

  If it is determined in step S706 that L is not equal to 0, the process proceeds to step S710, and has a character string of the Mth token among the child nodes of the node represented by the variable N, and the variable L The node information whose value is within the range between the maximum value and the minimum value of the length to the end node is substituted.

  In step S711, it is determined whether the value of the variable N is NULL. If the value of the variable N is NULL, there is no node corresponding to the Mth token. Therefore, the process proceeds to step S712, and it is determined that no character string is registered.

  If it is determined in step S711 that the value of the variable N is not NULL, the process returns to step S705.

  Note that the search device according to the present embodiment can be realized by causing a computer to execute a program. That is, it can be realized by realizing the index memory 101 as a computer memory or secondary storage, and causing the computer CPU to execute a program having the tokenizer 102 and the search unit 103 as modules.

  The module that implements the tokenizer 102, for example, reads a character string from an address of a memory in which a character string to be searched is stored, configures a data structure that represents a token string in an area in the memory, The information is transmitted to a module that implements the search unit 103.

  A module that implements the search unit 103 operates a register corresponding to the token string holding unit 602, the token number holding unit 603, the token number holding unit 604, and the node information holding unit 605, and the character string converted into a token string. Is present in the character string corresponding to the tree structure stored in the index memory 101.

  Note that the search device can be realized only by hardware by combining LSIs and the like.

(Example)
FIG. 8 shows an example of processing of the flowcharts of FIGS. As shown in FIG. 8A, if the registered character strings are “MR MAINET COLTD” and “MR MAINET SYSTEM COLT”, the processing shown in FIG. A tree structure is generated. Therefore, the index memory 101 stores tree structure data representing the tree structure shown in FIG.

  If the character string to be searched is “MR MAINET”, the tokenizer 102 generates a token string represented by the data structure shown in FIG. When this token sequence is transmitted to the search unit 103, the value of N becomes NULL in step S702. This is because node 1 has “MR”, which is the character string of the first token, but the number of tokens in the token sequence is 2, and 2 is the maximum length 4 from node 1 to the end node. And the minimum value 3 are not included, the process branches to step S704 in step S703, and it is determined that no character string is registered. Actually, (a) in FIG. 8 does not include a character string that matches “MR MAINET”. Therefore, it is possible to determine that “MR MAINET” is not registered without scanning the token corresponding to “MAINET”.

  If the search character string is “MR MAINET COLT”, the tokenizer 102 generates a token string represented by the data structure shown in FIG. When this token sequence is transmitted to the search unit 103, the values of the variables L, M, and N transition as shown in FIG. That is, the values of variables L, M, and N in step S701 are those shown in row 21. However, the value of N is before initialization and is set to NULL. When step S <b> 702 is executed, the values of variables L, M, and N are such that the value of N represents node 1, as shown in line 22. When step S705 ends, the values of variables L, M, and N are as shown in line 23. When step S706 is executed and step S710 is executed, the values of the variables L, M, and N are as shown in line 24. In step S711, since the value of N is not NULL, the process returns to step S705, and the values of variables L, M, and N are as shown in line 25.

  The values of variables L, M, and N are as shown in line 26 by the processing of steps S706 and S710. Since the value of N is not NULL in step S711, the processing returns to step S705, and the variables L, M, and N are returned. The value of N is as shown in line 27. In step S706, since the value of the variable L is 0, the process branches to step S707. Since node 3 is a terminal node, the process branches to step S709. Therefore, in step S709, it is determined that the character string is registered. In fact, in FIG. 8A, “MR MAINET coltd” exists in the second character string. In this way, even when searching for a registered character string, nodes that do not match the token string are not scanned.

(Application examples)
The search device according to the present embodiment detects, for example, whether or not the transfer addressee is registered as a business partner, and if the transfer addressee is a business partner, it is used to acquire the account number and the like. be able to. That is, in advance, each character string representing the name of the account holder of the customer's account is converted into a token string, and the tree structure data formed by merging the common token string from the start end is stored in the index memory 101. To do. At this time, the maximum value and the minimum value of the length to the end node are associated with the tree structure data.

  When the name of the addressee of the transfer is input, the character string representing the name is transmitted to the tokenizer 102, converted into a token string, and the character string representing the name by the search unit 103 is converted to the account of the customer. Detect if it exists in the string of the name of the holder.

  For example, when the processes of S708 and S712 in the flowchart of FIG. 7 are executed, it is assumed that the addressee of the transfer is not registered as a business partner. Further, by associating the account number of the business partner with the terminal node, when the process of step S709 is executed, the account number of the business partner is obtained by acquiring the account number associated with the node represented by the variable N. An account number or the like can be acquired, and a transfer process or the like can be performed.

(Embodiment 2)
FIG. 9 shows a modification of the data structure shown in FIG. 3 as Embodiment 2 of the present invention. 9A, the data structure has six slots, and the slots 901, 904, 905, and 906 correspond to the slots 301, 302, 303, and 304 in FIG. 3, respectively. In FIG. 9, slots 902 and 903 are shown. The slot 902 stores the first character string stored in the slot 901. The slot 903 stores the length of the character string. For example, when “MTU” is stored in the slot 901, M and 3 are stored in the slot 902 and the slot 903, respectively.

  FIG. 9B shows tree structure data corresponding to the character string shown in FIG. 8A using the data structure shown in FIG.

  When such a data structure is used, in step S702 and step S710 of the flowchart of FIG. 7, first, the first character of the token character string is the same as the character stored in the slot 902, and The length of the character string of the token is made to obtain the address of the same node as the value stored in the slot 903. If the first character of the token character string is the same as the character stored in the slot 902 and the length of the token character string is the same as the value stored in the slot 903, the node is the root node. Alternatively, if it does not exist as a child node, it is possible to detect that the character string to be searched does not exist in the plurality of character strings without comparing the character string of the token with the character string of the slot 901.

  Further, instead of associating the maximum value and the minimum value of the length to the end node with each node, a list of the length to the end node may be associated. In this case, if the number of unscanned tokens is not in the list associated with the child node, it can be detected that the character string to be searched does not exist in the plurality of character strings.

  100 search device, 101 index memory, 102 tokenizer, 103 search unit

Claims (8)

Each of a plurality of character strings is converted into one or a plurality of token strings and a common token string is merged from the start end, and the tree structure data nodes are connected to the end nodes. An index memory for storing tree-structured data associated with a maximum value and a minimum value;
A tokenizer that converts the string to be searched into a sequence of one or more tokens;
The character to be searched for is searched by selecting a child node of the node of the tree-structured data stored in the index memory while scanning the token row converted by the tokenizer and searching for a route starting from the root node. A search unit for detecting whether a column exists in the plurality of character strings,
The search unit searches for the character string to be searched if there is no child node including the number of unscanned tokens within the range of the maximum value and the minimum value of the length from the child node of the node to the end node. Is a search device for detecting that is not present in the plurality of character strings. The length of the character string of the token represented by the node is associated with the node constituting the tree structure data stored in the index memory,
If there is no child node associated with the length of the character string of the token to be scanned next from among the child nodes of the node, the search unit searches for the character string to be searched among the plurality of character strings. The search device according to claim 1, wherein the search device detects that it does not exist in the device. A node constituting the tree structure data stored in the index memory is associated with a list of lengths up to the end node,
If there is no child node including the number of unscanned tokens in the list associated with the child node among the child nodes of the node, the search unit searches for the plurality of character strings. The search device according to claim 1, wherein the search device detects that it is not present in the.   The search device according to claim 1, wherein the tokenizer converts the character string to be searched into a string of delimiter tokens according to a delimiter character in the character string to be searched.   5. The search device according to claim 1, wherein the tokenizer converts the character string to be searched into a string of tokens by dividing the character string into words according to the presence or absence of a predetermined word.   The search device according to claim 1, wherein the tokenizer sets one character in the character string to be searched as one token. Each of a plurality of character strings is converted into one or a plurality of token strings and a common token string is merged from the start end, and the tree structure data nodes are connected to the end nodes. In the computer storing the tree structure data in which the maximum value and the minimum value of the length are associated in the storage unit,
The search string is converted into a sequence of one or more tokens,
The character string to be searched is obtained by searching a path starting from a root node by selecting a child node of the node of the tree structure data stored in the index memory while scanning the converted token sequence. It is a program for detecting whether it exists in multiple character strings,
If there is no child node including the number of unscanned tokens within the range of the maximum value and the minimum value of the length from the child node of the node to the end node, the character string to be searched is the plurality of characters. A program that causes the computer to detect that it does not exist in a column. Each of a plurality of character strings is converted into one or a plurality of token strings and a common token string is merged from the start end, and the tree structure data nodes are connected to the end nodes. A computer operating method for storing tree structure data in which a maximum value and a minimum value of a length are associated with each other in a storage unit,
The calculator is
Convert the search string to a sequence of one or more tokens,
The character string to be searched is obtained by searching a path starting from a root node by selecting a child node of the node of the tree structure data stored in the index memory while scanning the converted token sequence. Detect if it exists in multiple strings,
If there is no child node including the number of unscanned tokens within the range of the maximum value and the minimum value of the length from the child node of the node to the end node, the character string to be searched is the plurality of characters. The way the computer works, including not being in a column.

Images