JP2010266946A - Device and method for inputting character string - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To input expressions configured of a plurality of word strings with high use frequency by a simple operation by using dictionary data which can be efficiently created. <P>SOLUTION: A storage part 400 stores "reading complementary dictionary data" configured on the basis of input reading and "spelling complementary dictionary data" configured on the basis of the spelling after conversion to be associated with each other as dictionary data prepared according to input contents on the basis of task content for which an input terminal 1 is used. Each data are configured like a tree structure with words having dependency relation as nodes. A control part 100 retrieves reading complementary dictionary data and spelling complementary dictionary data by using a retrieval key to be obtained from an input character string, and reads a conversion candidate node string from the spelling complementary dictionary data. When conversion candidates are read, the control part 100 reads the node string to which the same complementary unit number has been imparted as one node string in a batch from the complementary dictionary data. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a character string input device and a character string input method, and more particularly to a character string input device and a character string input method suitable for input work in a call center or the like.

  When inputting a character string as a natural language into a device such as a computer, the character string input from a keyboard or the like is generally converted into a natural language by performing input conversion according to the language type. ing. As such input conversion, a method of searching dictionary data using an input from a keyboard or the like as a key, extracting candidate words, displaying them on a screen, and inputting by selecting from among them is generally used. It is. However, this method displays all conversion candidates based on reading without considering anything before the character string being entered (determined character string). There is a disadvantage that the number of conversion candidates becomes very large. Therefore, there is a technique for displaying a word that is supposed to be input next as a conversion candidate using a relationship (frequency or the like) with a previously input word (for example, see Patent Document 1).

JP 2003-233605 A

  When a user uses a computer in a specific job, the same sentence or an expression such as an idiom may be input many times. However, in the method disclosed in Patent Document 1, a word that is supposed to be input next is displayed as a conversion candidate using a relationship (frequency, etc.) with a previously input word. Even for frequently used sentences and idioms, it is necessary to select and confirm words in order from the top, which is a complicated operation. For example, such inconvenience often becomes a problem in work of a call center where it is necessary to input the contents of reception by telephone in real time, and it is desired to input a character string more efficiently. Therefore, it is conceivable to register frequently used expressions in the dictionary and perform input conversion with a small number of conversions. In this case, the longer the word string constituting the expression, the greater the number of conversion candidates. For an input user, it does not necessarily lead to input efficiency. In addition, expressions that are frequently input often change over time, and it is necessary to update the dictionary data. However, expressions that consist of long word strings are also included for the dictionary creator. It is not realistic to build the dictionary data every time it needs to be updated.

  The present invention has been made in consideration of such circumstances, and its purpose is to input frequently used expressions composed of a plurality of words by a simple operation using dictionary data that can be efficiently created. Another object of the present invention is to provide a character string input device and a character string input method that can be used.

  In order to solve the above-described problem, the present invention provides a character string input device that searches for and presents conversion candidates according to an input character string and inputs a character string by selecting the presented conversion candidate. A dictionary in which readings before conversion processing of input character strings are recorded as words, and reading complementary dictionary data in which words having a dependency relationship with each other are configured in a tree structure as nodes, and A dictionary in which spelling after each word included in the reading completion dictionary data is converted as a word is recorded as a word, and the word before the conversion processing and the word after the conversion processing are associated as nodes and indicated by the reading completion dictionary data A storage unit configured to store a spelling completion dictionary data in which a same complementing unit number is assigned to a node having a dependency relationship, which is configured as a tree structure that matches the tree structure and is regarded as one node row; A search key generation unit that generates a search key based on an input; and the spelling corresponding to the search key by searching the reading completion dictionary data and the spelling completion dictionary data based on the search key generated by the search key generation unit The node of the complementary dictionary data is identified, and the spelling of the identified node and the spelling of the downstream node of the identified node assigned the same complement unit number as the identified node are read from the spelling complement dictionary data A conversion candidate extraction unit that acquires conversion candidates that are combined together, and a conversion candidate output unit that outputs the conversion candidates acquired by the conversion candidate extraction unit.

  Further, in order to solve the above-described problem, the present invention provides the above-described character string input device, wherein the conversion candidate extraction unit, when the search key includes unconfirmed hiragana, The most upstream node that completely matches the search key is detected, the spelling complement dictionary data node corresponding to the detected node and one downstream node are identified, and the spelling of the identified node and one downstream node is identified. And a conversion candidate obtained by combining the further downstream node spelling of the one downstream node assigned the same complement unit number as the one downstream node.

  Further, in order to solve the above-described problem, the present invention provides the character string input device described above, wherein the conversion candidate extraction unit uses the search key from the spelling completion dictionary data when the search key is a confirmed character. The most upstream node that is forward-matching and partial-matching, and the spelling of the identified node and the spelling of the downstream node of the identified node assigned the same complement unit number as the identified node. A combined conversion candidate is acquired.

  Further, in order to solve the above-described problem, the present invention provides the character string input device described above, wherein the conversion candidate extraction unit uses the search key from the spelling completion dictionary data when the search key is a confirmed character. The most upstream node that exactly matches the node, one node downstream of the identified node, the spelling of the identified node and one downstream node, and the same complement unit number as the one downstream node A conversion candidate obtained by combining the downstream node spelling of the one downstream node to which is assigned is acquired.

  In order to solve the above-described problem, the present invention provides the above-described character string input device, wherein the conversion candidate extraction unit includes an unconfirmed character string in which the search key includes a confirmed character and an unconfirmed hiragana. In this case, by using the fixed character of the search key, the most upstream node of the spelling completion dictionary data is searched to identify a matching node, and the downstream node of the reading completion dictionary data node corresponding to the identified node and the node It is determined that the search key matches the unconfirmed hiragana, and is determined from the spelling completion dictionary data to be the identified node and the downstream node of the identified node, and is determined to match the unconfirmed hiragana of the search key Read the spelling between the node corresponding to the node of the reading completion dictionary data and the further downstream node assigned the same completion unit number as the downstream node. Out, to obtain a conversion candidate bound the read spelling, characterized in that.

  Further, in order to solve the above-described problem, the present invention provides the above-described character string input device, which specifies a word string having an appearance frequency higher than a predetermined threshold from input sentence data, and the specified word A new node sequence generated by assigning the same complement unit number to the node sequence of the spelling complement dictionary data corresponding to the column is added to the spelling complement dictionary data, and the node sequence corresponding to the added node sequence is A dictionary generation unit for adding to the reading completion dictionary data is further provided.

  Further, in order to solve the above-described problem, the present invention provides the above-described character string input device, wherein each of the most upstream nodes of the spelling complement dictionary data has a node sequence starting from the most upstream node. If the number of acquired node strings is less than or equal to the predetermined number, the new node string generated by assigning the same complement unit number to the node string starting from the most upstream node is used as the spelling complement. In addition to adding to the dictionary data, a dictionary generation unit is further provided for adding a node sequence corresponding to the added node sequence to the reading completion dictionary data.

  In order to solve the above-described problem, the present invention searches and presents conversion candidates according to the input character string, and performs character string input by selecting the presented conversion candidate. A character string input method used in a device, wherein the input device is a dictionary in which readings before conversion processing of input character strings are recorded as words, and words that are in a dependency relationship with each other are represented as nodes. Reading dictionary data configured in a tree structure, and a dictionary in which spelling after each word included in the reading completion dictionary data is converted as a word is recorded before and after the conversion process. Associating words as nodes to form a tree structure that matches the tree structure indicated by the reading completion dictionary data, and assigning the same complementary unit number to the nodes in the dependency relationship that are regarded as one node string A search key generation process in which a search key generation unit of the character string input device generates a search key based on a user input, and conversion candidate extraction of the character string input device. The part searches the reading completion dictionary data and the spelling completion dictionary data based on the search key generated in the search key generation process, and identifies and specifies the node of the spelling completion dictionary data corresponding to the search key Conversion candidate extraction for acquiring a conversion candidate obtained by reading from the spelling complement dictionary data the spelling of a node and the spelling of a node downstream of the specified node assigned the same complement unit number as the specified node A conversion candidate output process in which the conversion candidate output unit of the character string input device outputs the conversion candidate acquired in the conversion candidate extraction process; Is a character string input method characterized by having a.

  According to the present invention, since a user can input and display an expression composed of a plurality of word strings as conversion candidates only by inputting the first character, it is possible to efficiently input a character string. In addition, it is possible to extract expressions that include frequently used expressions or words that are used only in limited expressions, and register them in the dictionary so that they can be handled as a single node sequence. This makes it possible to reduce the time and human resources required to build the dictionary.

It is a block diagram which shows the structure of the input terminal 1 by this embodiment. It is a block diagram which shows the structural example of the dictionary data by this embodiment. It is a schematic diagram which shows an example of the input screen by this embodiment. It is a conceptual diagram which shows the structural example of the reading complement dictionary data by this embodiment. It is a conceptual diagram which shows the structural example of the spelling complement dictionary data by this embodiment. It is a functional block diagram which shows the function implement | achieved by the control part 100 by this embodiment. It is a flowchart for demonstrating the operation | movement of a character string input process by this embodiment. It is a flowchart for demonstrating operation | movement of the search key production | generation process by this embodiment. It is a conceptual diagram which shows the various examples in a search key production | generation process. It is a flowchart for demonstrating operation | movement of the candidate extraction process by this embodiment. It is a flowchart for demonstrating operation | movement of the head unconfirmed process by this embodiment. It is a conceptual diagram which shows the various examples in a head unconfirmed process. It is a conceptual diagram which shows the various examples in a head unconfirmed process. It is a conceptual diagram which shows the various examples in a head unconfirmed process. It is a conceptual diagram which shows the various examples in a head unconfirmed process. It is a flowchart for demonstrating operation | movement of the head determination process by embodiment. It is a conceptual diagram which shows the various examples in a head determination process. It is a flowchart for demonstrating the operation | movement of the surplus matching process by this embodiment. It is a flowchart for demonstrating the operation | movement of the surplus matching process by this embodiment. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a conceptual diagram which shows the various examples in a surplus matching process. It is a flowchart for demonstrating the operation | movement of an output process by this embodiment. It is a conceptual diagram which shows the various examples in an output process. It is a schematic diagram for demonstrating the example of an output by this embodiment. It is a flowchart for demonstrating operation | movement of the supplement unit number assignment | providing process by this embodiment. It is a conceptual diagram which shows the example in the complement unit number provision process of FIG. It is a conceptual diagram which shows the example of a dictionary update in the supplement unit number provision process of FIG. It is a flowchart for demonstrating operation | movement of the supplement unit number assignment | providing process by this embodiment. It is a conceptual diagram which shows the example of a dictionary update in the complementary unit number provision process of FIG. It is a block diagram which shows the structural example at the time of applying this invention to a server client system.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the present embodiment, a case where the present invention is applied to an input terminal used for input work in a call center or the like will be described as an example.

The input terminal 1 according to the present embodiment is configured from an information processing apparatus such as a workstation or a personal computer, for example. Hereinafter, the input terminal 1 according to the present embodiment will be described with reference to FIG.
FIG. 1 is a schematic block diagram showing the configuration of the input terminal 1 according to the present embodiment. As illustrated, the input terminal 1 includes a control unit 100, an input control device 200, an output control device 300, a storage unit 400, and the like.

  The control unit 100 includes, for example, a CPU (Central Processing Unit), a memory serving as a work area, and the like. The control unit 100 reads out and executes a program stored in the storage unit 400, whereby the input terminal 1 Are controlled, and each process to be described later is executed.

  The input control device 200 connects, for example, an input device 210 such as a keyboard 211 or a pointing device 212 and the input terminal 1, and inputs an input signal corresponding to an operation of the input device 210 to the control unit 100. In the present embodiment, it is assumed that a character string is input using the keyboard 211 and an input area is selected using the pointing device 212.

  The output control device 300 connects, for example, the output device 310 such as the display device 311 and the input terminal 1, and outputs the processing result of the control unit 100 to the output device 310. In the present embodiment, an input screen prepared in advance is displayed by being output to the display device 311, and an input operation using the input device 210 is performed on the displayed input screen.

  The storage unit 400 includes a storage device such as a hard disk device, for example, and mainly stores information necessary for character string input processing according to the present embodiment. In the present embodiment, the storage unit 400 stores a dictionary storage area 410, a program storage area 440, a result table 450 that stores the searched conversion candidates, and a character string selected by the user using the input device 210 from the conversion candidates. It is assumed that a storage area such as an input sentence storage area 460 for storing the input sentence to be configured is secured.

  The dictionary storage area 410 stores dictionary data used when inputting a character string using the input terminal 1. In the present embodiment, it is assumed that Japanese character input as a natural language is performed in character string input using the input terminal 1. In this case, the input terminal 1 performs input conversion according to the input from the keyboard 211 by a so-called FEP (Front End Processor) operation. Therefore, the dictionary storage area 410 stores such dictionary data that is referred to during Japanese input conversion. In this case, the input terminal 1 refers to the dictionary data stored in the dictionary storage area 410 to display and output a conversion candidate (input candidate) character string in a selectable manner.

  An example of dictionary data according to the present embodiment will be described with reference to FIG. FIG. 2 is a diagram for explaining the configuration of dictionary data stored in the dictionary storage area 410. As shown in this figure, a basic dictionary area 420 and an input content-specific dictionary area 430 are secured in the dictionary storage area 410. The basic dictionary area 420 is an area for storing general dictionary data (hereinafter referred to as “basic dictionary data”) among the dictionary data referred to when inputting Japanese on the input terminal 1. In this basic dictionary data, various words are associated with information indicating parts of speech and readings. That is, general dictionary data used in a normal input FEP or the like is stored in the basic dictionary area 420 as basic dictionary data.

  The input content-specific dictionary area 430 is an area for storing dictionary data prepared according to the input content based on the business content used by the input terminal 1. For example, when the input terminal 1 is used for call center operations, the input items are defined in advance, so that dictionary data composed of word groups used for input with specific input items is stored in the input content-specific dictionary region 430. The That is, dedicated dictionary data prepared for each input content is stored in the dictionary region 430 for each input content.

  As an example, reception input dictionary data (A1) 431, address input dictionary data (A2) 432, and the like as shown in FIG. Such input content-specific dictionary data is associated with the input area of the input screen. Here, FIG. 3 is a diagram illustrating an example of the input screen. For example, in the case of the input screen IS as shown in this figure, the reception input dictionary data 431 is associated with the input area A1 for inputting the content of the telephone reception. The address input dictionary data 432 is associated with an input area A2 for inputting address information. That is, on the input screen IS as shown in FIG. 3, the contents input to the input area are defined for the input area associated with the dictionary data.

  Each of the dictionary data prepared for each input content includes “reading dictionary data” configured based on the input reading and “spelling dictionary data” configured based on the spelling after conversion. Are included as a pair. That is, as shown in FIG. 2, the reception input dictionary data (A1) 431 includes the reading complement dictionary data 431A and the spelling complement dictionary data 431B in association with each other. The address input dictionary data 432 (A2) includes reading complement dictionary data 432A and spelling complement dictionary data 432B in association with each other.

  Each of the dictionary data prepared for each input content is composed of such a pair of different dictionary data, and each of the response input dictionary data (A1) 431 and the address input dictionary data (A2) 432 is a set. Dictionary data. Therefore, if only the response input dictionary data (A1) 431 and the address input dictionary data (A2) 432 are stored in the input content-specific dictionary region 430, the input content-specific dictionary region 430 includes two sets. Dictionary data is stored.

  The configuration of such input content-specific dictionary data will be described with reference to FIGS. 4 and 5. FIG. Here, the configuration of the response input dictionary data 431 will be described as an example of input content dictionary data. FIG. 4 is a conceptual diagram illustrating a configuration example of the reading complement dictionary data 431A, and FIG. 5 is a conceptual diagram illustrating a configuration example of the spelling complement dictionary data 431B. In the present embodiment, a case will be described in which the response input dictionary data 431 is configured based on a vocabulary that frequently appears at the time of a telephone response to a user in a call center that performs a user support service for a mobile phone.

With reference to FIG. 4, a configuration example of the reading completion dictionary data 431A will be described.
As shown in the figure, the reading complement dictionary data 431A is composed of binary data having a tree-like hierarchical structure (tree structure) with “reading” of a word as a dictionary element as a node (node). This hierarchical structure is a configuration (node string) in which root is the apex, and nodes in the dependency-side clause of the dependency relationship are sequentially connected to nodes in the receiver-side clause. Here, in the node sequence, a node closer to the root is called an upstream node, and a node far from the root is called a downstream node. Therefore, one downstream of the root is the most upstream node. With this configuration, a dictionary is constructed from data indicating a state in which words that are grammatically dependent on each other are linked at relative positions. Note that the number of nodes constituting one clause may be one or more. Here, since the reading complement dictionary data 431A is a dictionary based on “reading” of the input word, each node is indicated with a headword by hiragana representing reading. In FIG. 4, the reading of each node of the reading complement dictionary data 431A is shown in <>.

  FIG. 5 shows a configuration example of the spelling complement dictionary data 431B paired with such reading complement dictionary data 431A. As shown in the figure, the spelling complement dictionary data 431B is configured by the same hierarchical structure (tree structure) as the reading complement dictionary data 431A, and “spelling” corresponding to the readings indicated in each node of the reading complement dictionary data 431A. Is configured as a headword of each node of the spelling complement dictionary data 431B at the position of the node of the spelling complement dictionary data 431B corresponding to the node of the reading complement dictionary data 431A. In FIG. 5, the spelling of each node in the spelling complement dictionary data 431B is shown in <>.

  That is, the spelling complement dictionary data 431B is dictionary data in which each node included in the reading complement dictionary data 431A is configured based on the spelling of the node. For example, in the reading completion dictionary data 431A, the node in the spelling completion dictionary data 431B linked to the node whose reading is “Keitai” corresponds to the post-conversion notation of “Keitai”, that is, the spelling “mobile”. To do.

  Here, the symbol in [] attached to each node of the reading complement dictionary data 431A and the spelling complement dictionary data 431B is a link pointer representing a link between the nodes of the reading complement dictionary data 431A and the spelling complement dictionary data 431B. It is. That is, between the reading complement dictionary data 431A and the spelling complement dictionary data 431B, nodes having a link pointer of the same symbol in [] can be referred to each other. Here, “′” is added to the symbol of the link pointer to distinguish the link pointer of the reading completion dictionary data 431A. That is, if the link pointer given to the node of the spelling complement dictionary data 431B is [xn], the link pointer of the corresponding node of the reading complement dictionary data 431A is [xn ′]. As a result, the pair of reading completion dictionary data 431A and the spelling completion dictionary data 431B are linked in units of nodes.

  Hereinafter, the link pointer xn ′ and the node whose reading is “**” will be described as the node xn ′ “**” or simply the node xn ′, and the node whose link pointer xn and the spelling is “**” will be referred to as the node xn “ ** "or simply node xn. For example, the node of the link pointer a1 ′ in the reading completion dictionary data 431A is described as the node a1 ′ “Keitai” or the node a1 ′, and the node of the link pointer a1 in the spelling completion dictionary data 431B is the node a1 “mobile” or the node a1. It describes.

  The numerical values in parentheses attached to the nodes of the reading complement dictionary data 431A and the spelling complement dictionary data 431B are the complement unit numbers, and the corresponding node of the reading complement dictionary data 431A by the link pointer and the spelling complement dictionary The same complement unit number is described in the node of the data 431B. Adjacent nodes adjacent to each other with the same numerical complement unit number are regarded as one node string. Hereinafter, from a certain node xn (or node xn ′), the same complement unit number as that of the node xn (xn ′) is given, and a node sequence consisting of continuous nodes downstream from the node xn (xn ′) It is described as the same complement unit node string of the node xn (xn ′). The same complement unit number is assigned to at least the nodes constituting the same phrase. The same complement unit node string may be composed of one node.

  For example, since the node a1 “mobile” in the spelling complement dictionary data 431B has a complementing unit number of 1 and the node a2 “no” has a complementing unit number of 2, it is regarded as a node included in a different same complementing unit node sequence. On the other hand, the node i1 “mobile”, the node i2 “no”, and the node i3 “screen” have the complementary unit number 1, so the nodes i1, i2, and i3 are the same complementary unit node sequence and are regarded as one node sequence. It is.

  The numerical value in << >> attached to each node of the reading complement dictionary data 431A and the spelling complement dictionary data 431B indicates the number of appearances in the sentence actually input to the input area. For example, when the number of appearances of “new” is 50, the number of appearances 50 is set in the nodes f1 and j1 “new” of the spelling completion dictionary data 431B, and “new / mobile / buy / buy” (“/” Indicates the position where the nodes are combined), the number of appearances 5 is set in the node f4 “Bought”. Instead of the number of appearances, arbitrary information indicating the appearance frequency, such as the rate of appearance of words and the standard deviation, can be used.

  In the above, the dictionary data has a hierarchical structure with nodes that constitute clauses of three or less levels, but may have a hierarchical structure with nodes that constitute clauses of two layers, or a node that has four or more layers as the maximum clause. A hierarchical structure may be used.

  The program executed by the control unit 100 is stored in the program storage area 440 shown in FIG. And the function at the time of performing the character string input process concerning this embodiment is implement | achieved because the control part 100 runs the program stored in the program storage area 440. FIG. Functions realized by the control unit 100 are shown in FIG.

  FIG. 6 is a functional block diagram illustrating functions realized by the control unit 100. As illustrated, the control unit 100 includes functions of a dictionary selection unit 110, a search key generation unit 120, a conversion candidate extraction unit 130, a conversion candidate output unit 140, a character string input unit 150, a dictionary generation unit 160, and the like.

  The dictionary selection unit 110 determines which of the input areas on the input screen IS is input, and selects a dictionary corresponding to the input area from the dictionary storage area 410 based on the determined result.

  The search key generation unit 120 acquires a character string from the input area where the input is performed, generates a search key, and delivers it to the conversion candidate extraction unit 130.

  The conversion candidate extraction unit 130 extracts conversion candidate words based on the search key generated by the search key generation unit 120 and the dictionary specified by the dictionary selection unit 110 and stores the conversion candidate words in the result table 450 of the storage unit 400.

  The conversion candidate output unit 140 deletes redundant portions from the candidate words extracted by the conversion candidate extraction unit 130, as will be described later. Then, in cooperation with the output control device 300, the formatted character string is displayed on the display device 311 so that it can be selected.

  The character string input unit 150 is a character selected by the input device 210 among the formatted character strings displayed and output on the display device 311 so as to be selectable by the conversion candidate output unit 140 in cooperation with the input control device 200 or the like. The string is input as a converted character string corresponding to the unconfirmed character string or a character string following the confirmed character string. In this case, in cooperation with the output control device 300, the selected character string is displayed in the input area where the input operation is currently being performed, and a sentence composed of the selected character string is displayed in the input sentence storage area 460. Write to. That is, the text actually input in each input area is written in the storage unit 400.

  The dictionary generation unit 160 detects a word string having an appearance frequency higher than a predetermined value from a sentence in each input area written in the input sentence storage area 460 and a word string with few dependency branches in the dictionary data. In the reading completion dictionary data and spelling completion dictionary data corresponding to the area, a node string indicating the detected word string is identified, and a new node string generated by assigning the same complement number to the identified node string is added.

  In the present embodiment, the functional configuration shown in FIG. 6 is logically realized by the control unit 100 executing a program. These functions are, for example, integrated in an application specific integrated circuit (ASIC). It may be realized by a physical configuration such as a circuit).

  Next, the operation of the input terminal 1 having the above-described configuration will be described. Here, “character string input processing” executed when a character string is input using the input terminal 1 will be described with reference to FIGS. 7 to 28. For example, this process is started when an operation for displaying the input screen IS as shown in FIG. 3 is performed.

  FIG. 7 is a flowchart of the character string input process in the input terminal 1. In the same figure, when the processing is started and an input is made from the input device 210 by the operator, the dictionary selection unit 110 based on the input signal from the input control device 200 to the control unit 100, It is determined which input area is designated (S001). That is, it is determined whether or not the input area to be input is selected from the input areas prepared on the input screen IS by the operation of the pointing device 212 or the like by the operator of the input terminal 1.

  When any input area is designated (S001: YES), the dictionary selection unit 110 determines whether or not the designated input area is a specific input area (S002). In the present embodiment, an input area that is associated with dictionary data and that is a target of conversion input using a dictionary according to input content is set as a specific input area. Here, there are an input area A1 for inputting the response contents and an input area A2 (see FIG. 3) for inputting an address.

  When it is determined that the designated input area is not a specific input area (S002: NO), the basic dictionary data in the basic dictionary area 420 is referred to, and a character string is input by a normal input conversion operation (S100). ).

  On the other hand, when it is determined that a specific input area has been designated (S002: YES), the dictionary selection unit 110 detects the type of the designated input area. Here, when the input area A1 is designated, the identifier “A1” associated with the input area is detected, and when the input area A2 is designated, the identifier is associated with the input area. The identifier “A2” is detected.

  When the dictionary selection unit 110 detects the identifier, the dictionary selection unit 110 accesses the dictionary area 430 for each input content, and selects dictionary data corresponding to the detected identifier (S003). As shown in FIG. 2, identifiers indicating corresponding input areas (“A1” and “A2” in the example of FIG. 2) are included in a plurality of sets of input content dictionary data stored in the input content dictionary area 430. Is granted. Thus, the dictionary selection unit 110 selects dictionary data to which a corresponding identifier is assigned based on the detected identifier. Hereinafter, in the present embodiment, description will be made assuming that a character string is input to the input area A1. Therefore, the response input dictionary data (A1) 431 is selected here.

  When the dictionary data to be used is selected, the dictionary selection unit 110 notifies the conversion candidate extraction unit 130 of the selected dictionary data and also notifies the search key generation unit 120 of the designated input area.

  When receiving the notification of the input area from the dictionary selection unit 110, the search key generation unit 120 acquires a character string in the designated input area and generates a search key (S200). Various methods are conceivable for generating the search key. In this embodiment, the search key is generated according to the flowchart shown in FIG.

  When the conversion candidate extraction unit 130 receives the search key generated in S200 from the search key generation unit 120, the conversion candidate extraction unit 130 extracts conversion candidates from the dictionary selected by the dictionary selection unit 110 in S003 using the search key, and obtains a search result. Output (S300). If a candidate word as a search result is not extracted (S004: NO), the conversion candidate extraction unit 130 generates a search key shown in FIG. If the extraction of the candidate word is successful (S004: YES), the conversion candidate output unit 140 performs shaping for displaying the search result output in S300 as a candidate (S400).

  After S400, when input by the input device 210 is not performed (S005: NO), or when characters are further input in the currently specified input area (S005: YES, S006: YES), The process from S200 is performed. On the other hand, when the input by the input device 210 is neither a character input to the currently designated input area nor an input of an end instruction (S006: NO, S007: NO), the process from S001 is performed, and an end instruction is issued. If it is an input (S006: NO, S007: YES), the process is terminated.

Next, the operation of the search key generation unit 120 will be described with reference to FIGS. FIG. 8 shows a flowchart of the search key generation process executed by the search key generation unit 120 in S200 of FIG. 7, and FIG. 9 shows an example of the search key generation process.
When the search key generation process is executed in S200 of FIG. 7, in FIG. 8, the search key generation unit 120 first acquires a character string in the designated input area (S201). In general, a fixed character string that has been converted and an undetermined character string that is being input for conversion are mixed in the input area. The unconfirmed character string includes a “Hiragana” state before the FEP functions, and a state in which conversion candidates such as kanji are temporarily displayed by the FEP function. Hereinafter, the undetermined character string portion is indicated by a character surrounded by “” in the specification and by underlining in the drawing. An example of “confirmed character string + unconfirmed character string” is as shown in FIG. Here, when inputting the character string of “mobile phone”, the Chinese character of the character string of “mobile” is fixed, and the “phone” part is an unconfirmed character string before the FEP functions. Then, the character string “Denwa” is temporarily converted as “telephone”.

  Next, the search key generating unit 120 deletes the character string before the last punctuation “.” From the obtained character string (S202). An example is shown in FIG. Here, delete the character string "User is mobile. Broken" from the string "User is mobile. Broken mobile phone. Replace shop" anywhere in the input area, and press the search key. "Everywhere shop" to "exchange".

  Next, when the unconfirmed part is included in the search key and there is a confirmed character string on the right side of the unconfirmed part, it is deleted (excluded) from the search key (S203). An example is shown in FIG. If the search key generated in S202 is “any shop“ to ”exchange”, “exchange” is deleted and the search key “any shop“ to ”” is used.

In general, there are never a plurality of undetermined parts in a character string. Therefore, the state of the character string in the input area is limited to the following cases.
(1) Only the confirmed part (2) Confirmed part + Unconfirmed part (3) Confirmed part + Unconfirmed part + Confirmed part (4) Only unconfirmed part (5) Unconfirmed part + Confirmed part

  Of these, as a result of the processing of S203, the case of (3) results in the case of (2), and the case of (5) results in the case of (4). , (2) “determined part + undetermined part”, (4) “only undetermined part”.

Next, the search key generation unit 120 separates the search key into a “determined part” and an “unconfirmed part” (S204), and acquires the length L of the confirmed part (S205). For example, when the deterministic part is “where shop”, L = 7. Next, the search key generation unit 120 extracts a character string having a length L from the end (right end) to the left side of the confirmation unit, and generates a “end part” (S206). In the first search key generation, the end part is equal to the length of the fixed part. Therefore, the tail part generated in S206 is “where shop”.
Finally, the search key generation unit 120 combines the “end part” and the “undefined part” (S207). As a result, in the above example, the character string shown in FIG. 9D is output as a search key. That is, in S203, when the search key “any shop” is extracted and the end part is “where shop”, the end part “where shop” and the unconfirmed part “also” are combined. The search key is “anywhere”. In the case of (1), since there is no “undetermined part”, only “tail part” is the search key. When the search key is generated in this way, the process of S300 in FIG. 7 is subsequently performed.

In FIG. 7, when a candidate word is not extracted (S004: NO), the search key generation unit 120 determines whether L = 0 (S209). If L is 0 (S209: YES), the process of S400 in FIG. 7 is executed. If L is not 0 (S209: NO), the current L is updated to a value obtained by subtracting 1 and the left end of the determination unit is displayed. One character is reduced, the process from S206 is performed, and after the execution of S207, the process of S300 in FIG. 7 is performed.
For example, when a candidate word is not extracted when L = 7 and the last part is “where shop”, in S206, (L−1) characters are extracted from the right end, which is the “end” of the final part, and the end Since L = 6, the last part is “this shop”. In S207, the search key is “this shop“ also ””.

  Next, the operation of the conversion candidate extraction unit 130 will be described with reference to FIGS.

FIG. 10 shows a flowchart of candidate extraction processing executed by the conversion candidate extraction unit 130 in S300 of FIG.
In FIG. 10, the conversion candidate extraction unit 130 first determines whether or not the head of the search key is an “unconfirmed unit” consisting of “Hiragana only” (S301). If the head of the search key is “Hiragana only” and “unconfirmed part” (S301: YES), the process proceeds to “Head unconfirmed process” (S500), and “Hiragana only” is formed. If it is not a “determining part” (S301: NO), the process proceeds to the “leading determination process” (S600).

Next, the “start unconfirmed process” in S500 in the candidate extraction process shown in FIG. 10 will be described with reference to FIGS. FIG. 11 shows a flowchart of the head unconfirmed process executed by the conversion candidate extraction unit 130 in S500 of FIG. 10, and FIGS. 12 to 15 show examples of the head unconfirmed process.
In FIG. 11, the conversion candidate extraction unit 130 first acquires one most upstream node that is one node downstream from the “reading supplement dictionary” (S501). This is designated as node x. Hereinafter, when “node x” is described using lower case alphabet letters, it represents a node of the reading completion dictionary, and when “node X” is written using upper case letters, it corresponds to node x. It represents a spelling completion dictionary node.

When there is some “forward match” between the node x and the search key, the inclusion relation is defined as follows according to the length of the search key and the node x.
Search key ⊂ Node x Partial match Search key = Node x Exact match Search key ⊃ Node x Extra match

Further, the reading complement dictionary and the spelling complement dictionary are configured as shown in FIGS. 4 and 5, respectively. An example in which the reading of the node x is “Keitai” is an example of partial match, complete match, and surplus match with the search key, as shown in FIG. In other words, if the search key is an unconfirmed part “Kei”, it is a partial match. is there.
It should be noted that the forward match means that the match of the head part (the left end of the character string) is always a condition when determining the match between the search key and the node x. FIG. 12B shows an example of partial matching but not forward matching and an example of surplus matching but not forward matching. That is, if the search key is the unconfirmed part “tai”, it is a partial match but not a forward match. Also, if the search key is the unconfirmed part “I want to”, it is a surplus match, but not a forward match.

If there is no match (forward match) between the search key and the node x (S502: NO), the conversion candidate extraction unit 130 checks whether there are any remaining nodes to be compared in the reading completion dictionary (S506). If there remains a node to be compared (S506: NO), the process returns to S501, and one node downstream of root that has not been compared is selected, and the subsequent processing is performed with this node as node x.
On the other hand, since all nodes have been compared, if there are no remaining nodes to be compared (S506: YES), the head unconfirmed process ends.

Further, if the search key and the node x match forward in any of partial match, complete match, and surplus match (S502: YES), the conversion candidate extraction unit 130 uses the link pointer assigned to the node x. The node X of the spelling complement dictionary is acquired (S503).
If the search key and the node x partially match (S504: YES), the conversion candidate extraction unit 130 reads the spelling of the same complement unit node string of the node X from the spelling complement dictionary and adds it to the result table 450 (S505). ). At this time, the conversion candidate extraction unit 130 reads the number of appearances of the last node of the same complement unit node sequence from the spelling complement dictionary, and writes it in the result table 450 in association with the combined spelling. An example of this is shown in FIG. However, in FIG. 13, the number of appearances written in the result table is omitted.

  As shown in FIG. 13A, for example, when the search key is the unconfirmed character “Kei” and the node x is the node a1 ′ “Keitai” in the reading complement dictionary data 431A, the spelling is partially matched. The node a1 “mobile” in the complementary dictionary data 431B is identified as the node X. The node a1 “mobile” is assigned the complement unit number “1”, and the node a2 “no”, the node c2 “ga”,... The numbers are different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, and adds the spelling “mobile” and the number of appearances of the node a1 to the result table 450 in association with each other.

  In addition, when the search key is the unconfirmed character “Kei” and the node x is the node d1 ′ “Keitai Denwa” in the reading completion dictionary data 431A, it partially matches, so the node d1 “of the spelling completion dictionary data 431B”. “Mobile phone” is identified as node X. A complement unit number “1” is assigned to the node d1 “mobile phone”, a complement unit number “2” is assigned to the node d2 “no” one downstream, and the complement unit numbers are different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, associates the spelling “mobile phone” of the node d1 and the number of appearances, and adds it to the result table 450.

  In addition, when the search key is the unconfirmed character “Kei” and the node x is the node i1 ′ “Keitai” in the reading completion dictionary data 431A, it partially matches, so the node i1 “mobile” in the spelling completion dictionary data 431B. Is identified as node X. The node i1 “mobile” is assigned the complementary unit number “1”, and the downstream node i2 “no” and the node i3 “screen” are also assigned the same complementary unit number “1”. Therefore, the conversion candidate extraction unit 130 associates the node i1 + node i2 + node i3 with the same complement unit node string and associates the spelling of the nodes i1, i2, and i3 with the number of appearances of the node i3. To the result table 450.

  On the other hand, as shown in FIG. 13 (b), if the search key is the unconfirmed character “A” and the node x is the node f1 ′ “new” in the reading completion dictionary data 431A, it is partially matched. The node f1 “new” of the dictionary data 431B is specified as the node X. The node f1 “new” is assigned the complement unit number “1”, and the downstream node f2 “mobile”, the node h2 “computer”,. Are different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, and adds the spelling “new” and the number of appearances of the node f1 to the result table 450 in association with each other.

  In addition, when the search key is the undetermined character “A” and the node x is the node j1 ′ “new” in the reading complement dictionary data 431A, partial matching is made, so the node j1 “new” in the spelling complement dictionary data 431B is Identified as node X. A complement unit number “1” is assigned to the node j1, and a complement unit number “2” is assigned to the downstream node j2 “PC”, and the complement unit numbers are different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, and adds the spelling “new” and the number of appearances of the node j1 to the result table 450 in association with each other.

  Thereafter, if there are still candidates to be compared (S506: NO), the process returns to S501. If the comparison has been completed for all the nodes x (S506: YES), the head unconfirmed process is terminated.

  On the other hand, when the search key and the node x completely match (S504: NO, S507: YES), the conversion candidate extraction unit 130 performs the same complement of the node X and the node immediately downstream of the node X from the spelling complement dictionary. The spelling of the unit node sequence is read and combined, and added to the result table 450 (S508). At this time, the conversion candidate extraction unit 130 reads the number of appearances added to the last node of the same complement unit node sequence from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling. An example of this is shown in FIG. However, in FIG. 14, the number of appearances written in the result table is omitted.

  As shown in FIG. 14A, for example, when the search key is the unconfirmed character “Keitai” and the node x is the node a1 ′ “Keitai” in the reading completion dictionary data 431A, the match is complete. The node a1 “mobile” in the spelling complement dictionary data 431B is identified as the node X. The conversion candidate extraction unit 130 reads the node a2 “no”, the node c2 “ga”,..., And the complement unit number “2” that are one downstream of the node a1. Then, for each of the read downstream nodes, a node string obtained by combining further downstream nodes to which the same complement unit number is assigned, node a2 “no” + node a3 “screen”, node a2 “no” + node b3 “ “Is there”, node c2 “ga” + node c3 “broken”,... Therefore, the conversion candidate extraction unit 130 corresponds to (1) the “portable screen” in which the spelling of the node a1 (node X) and the node a2 + node a3 (same complement unit node string) is combined, and the number of appearances of the node a3. (2) Correspondence between “there is a mobile phone” that combines the spelling of the node a1 (node X) and the node a2 + node b3 (same complement unit node string) and the number of appearances of the node b3, (3) node The association between the spelling of a1 (node X) and node c2 + node c3 (same complement unit node string) and the number of appearances of node c3 are added to the result table 450.

  Further, when the search key is the unconfirmed character “Keitai” and the node x is the node i1 ′ “Keitai” in the reading completion dictionary data 431A, it is completely matched. Therefore, the node i1 “mobile” in the spelling completion dictionary data 431B Is identified as node X. The conversion candidate extraction unit 130 reads the node i2 “no” one downstream of the node i1 and its complement unit number “1”. Then, for the read downstream node, a node string obtained by combining further downstream nodes to which the same complement unit number is assigned, node i2 “no” + node i3 “screen”... Therefore, the conversion candidate extraction unit 130 associates the spelling of the node i1 (node X) and the node i2 + node i3 (same complement unit node string) with the number of appearances of the node i3, and the result. Add to table 450.

  On the other hand, as shown in FIG. 14B, for example, when the search key is the unconfirmed character “sieving” and the node x is the node k1 ′ “sieving” in the reading complement dictionary data 431A, The node k1 “old” of the spelling complement dictionary data 431B is identified as the node X. A node k2 “mobile” that is one downstream of the node k1 is assigned a complement unit number “1”, and a node k3 that is one downstream of the node k2 is assigned a complement unit number “2”. Are different. Therefore, the conversion candidate extraction unit 130 sets only the node k2 as the same complement unit node string, and combines the spelling of the node k1 (node X) and the node k2 (same complement unit node string), and the node k2 Is added to the result table 450.

  On the other hand, as shown in FIG. 14 (c), when the search key is the unconfirmed character “new” and the node x is the node f1 ′ “new” (node x) in the reading complement dictionary data 431A, they completely match. Therefore, the node f1 “new” of the spelling complement dictionary data 431B is specified as the node X. A complement unit number “2” is assigned to the node f2 “mobile”, the node h2 “personal computer”,... One downstream of the node f1, and a complement unit number “3” is assigned to the downstream nodes f3, g3, and h3. The supplementary unit number is different. Therefore, the conversion candidate extraction unit 130 (1) associates the spelling of the node f1 (node X) with the spelling of the node f2 (same complementary unit node string) and the number of appearances of the node f2, ( 2) A “new personal computer” in which the spelling of the node f1 (node X) and the node h2 (same complementary unit node string) is combined, and the association of the number of appearances of the node h2, etc. are added to the result table 450.

  Further, when the search key is the unconfirmed character “new” and the node x is the node j1 ′ “new” in the reading completion dictionary data 431A, the search key is completely matched. Therefore, the node j1 of the spelling completion dictionary data 431B as the node X “New” is identified. A complement unit number “2” is assigned to the node j2 “personal computer” that is one downstream of the node j1, and a complement unit number “2” is also assigned to the downstream node j3 “no” and the node j4 “release date”. Has been granted. Accordingly, the conversion candidate extraction unit 130 combines the spelling of the same complement unit node sequence consisting of the node j1 (node X) and the node j2 + node j3 + node j4, and the number of appearances of the node j4. Are added to the result table 450 in association with each other.

  Thereafter, if there are still candidates to be compared (S506: NO), the process returns to S501. If the comparison has been completed for all the nodes x (S506: YES), the head unconfirmed process is terminated.

If the search key and the node x are neither partially matched nor completely matched, a surplus match is obtained (S507: NO). Therefore, first, the conversion candidate extracting unit 130 and substituting 1 to n (S509), to obtain one downstream node _{y n} nodes from reading complement dictionary _{y n-1} (S510). However, since the node _{y 0} is node x, in the first loop, a node _y n (n = 1) is a one downstream node of the node x.

Next, the conversion candidate extraction unit 130 generates a surplus key that is a remainder obtained by deleting the node x from the search key when the surplus key has not yet been generated, and if the surplus key has already been generated, the current surplus key is generated. Delete the node y _n-1 from the key and update the surplus key. Conversion candidate extracting unit 130, surplus key to node y _n, (or partial match or exact) is or includes a forward match condition is determined (S511). Therefore, in the first loop, it is determined whether or not the surplus key obtained by deleting the node x from the search key is included in the node y ₁ that is one downstream of the node x.

When the surplus key is included in a node _{y n} (S511: YES), the same node _{y n} and identifies the node _{Y n} corresponding spelling complement dictionary, the nodes from the node _{Y 1} corresponding to the node _{y 1 Y n} The spelling up to the complement unit node row is acquired as the node row Y (S512). Thus, in the first loop, the node string Y is the same complementary units node sequence of the node Y _1. The conversion candidate extraction unit 130 adds the spelling of the node X + node string Y to the result table 450 (S513). At this time, the conversion candidate extraction unit 130 reads the number of appearances of the last node of the node string Y from the spelling complement dictionary, and writes it in the result table 450 in association with the combined spelling.

On the other hand, in S511, if the surplus key is determined not included in node _{y n} (S511: NO), the conversion candidate extraction unit 130, or the node _{y n} is included in excess keys forward match condition, i.e., excess key to determine whether the excess matched against nodes _{y n} (S515). If the surplus keys to excess matched against nodes _{y n (S515: YES),} 1 is added to the value of the current n (S516), the processes from S510.
For example, in S516, when the value of n is updated to 2, in S510, the conversion candidate extracting unit 130 acquires one downstream node _{y 2} nodes _{y 1} from the complementary dictionary readings. Then, in S511, the conversion candidate extraction unit 130, the remainder obtained by deleting the node y ₁ from the current surplus key update surplus key update surplus key to node y ₂ was, determined either included forward matching condition To do.

After the processing of S513, or if the surplus key in S515 is determined not to excess matched against nodes _{y n} (S515: NO), if the still downstream node candidate to be compared (S514: NO) , Return to S509. On the other hand, if the comparison has been completed for all the nodes y ₁ (S514: YES), the process proceeds to S506.

  FIGS. 15A, 15B, and 15C show examples in which the remainder indicated by the surplus key is a partial match, complete match, and surplus match. However, in FIG. 15, the number of appearances written in the result table is omitted.

As shown in FIG. 15A, for example, when the search key is the unconfirmed character “Keitai” and the node x is the node a1 ′ “Keitai” (node x) in the reading complement dictionary data 431A. Matches surplus. Conversion candidate extraction unit 130, the node a1 '1 next downstream node a2' of "no", the node c2 'read from the "ga" complementary dictionary data 431A read, the node _{y 1.} Since the surplus key is “NO”, the conversion candidate extraction unit 130 specifies the reading “NO” of the completely matched node y ₁ . Then, by the link pointer of the node a2 'is a node y ₁ that this particular, from spelling supplement dictionary data 431 B, node a2 "no" is specified. The conversion candidate extraction unit 130 reads the node a2 “no” + node a3 “screen” and the node a2 “no” + node b3 “whether” as the same complement unit node row of the node a2, and sets it as the node row Y. The conversion candidate extraction unit 130 associates (1) the “portable screen” in which the spelling of the node a1 (node X) and the node a2 + node a3 (node row Y) is combined, and the number of appearances of the node a3, (2) Add the correspondence between “there is a mobile phone” that combines the spelling of the node a1 (node X) and the node a2 + node b3 (node row Y) and the appearance count of the node b3 to the result table 450 .
Further, even when the node x is the node i1 ′ “Keitai” (node x) in the reading completion dictionary data 431A, the node y ₁ (node i2 ′ “no”) and the surplus key “no” completely match, and the node i1 “Portable screen” in which the spelling of (node X) and node i 2 + node i 3 (node row Y) is combined, and the association of the number of appearances of node i 3 are added to result table 450.

On the other hand, as shown in FIG. 15B, when the search key is the undetermined character “new path” and the node x is the node f1 ′ “new” in the reading complement dictionary data 431A, there is a redundant match. Conversion candidate extracting unit 130, a node f1 'one of the downstream node f2'"form", h2 'read from the complementary dictionary data 431A Read "computer", and the node _{y 1.} Since the surplus key is “Pa”, the conversion candidate extraction unit 130 specifies the reading “Pason” of the node y _{1 that} matches forward and partially matches. Then, by the link pointer of the specified node y ₁ a is a node h2 ', in spelling supplement dictionary data 431 B, identifies the node h2 "PC". A complement unit number “2” is assigned to the node h2 “personal computer”, a complement unit number “3” is assigned to the node h3 that is one downstream of the node h2, and the complement unit number is different. Therefore, the conversion candidate extraction unit 130 sets only the node h2 as the same complement unit node string, and sets the same complement unit node string of the node h2 as the node string Y. The conversion candidate extraction unit 130 adds the “new personal computer” in which the spelling of the node f1 (node X) and the node h2 (node string Y) is combined, and the association of the number of appearances of the node h2 to the result table 450.

In addition, when the search key is the undetermined character “new path” and the node x is the node j1 ′ “new” in the reading complement dictionary data 431A, there is a redundant match. Conversion candidate extraction unit 130 reads from the complementary dictionary data 431A Read "computer"'one downstream node j2' of nodes j1, and the node _{y 1.} Since the surplus key is “Pa”, it coincides with node y ₁ in a forward match and a partial match. Therefore, the conversion candidate extracting unit 130, by the link pointer of the node j2 'is a node y _1, the spelling supplement dictionary data 431 B, a node j2 specifies "computer", as the same complementary units node sequence of the node j2, node j2 “Personal computer” + node j3 “no” + node j4 “release date” is read out and set as node string Y. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node j2 + node j3 + node j4 (node string Y) and the number of appearances of the node j4. It adds to the result table 450.

Further, as shown in FIG. 15C, when the search key is the unconfirmed character “I want to wake up” and the node x is the node f1 ′ “new” in the reading complement dictionary data 431A, there is a surplus match. Conversion candidate extracting unit 130, a node f1 'one of the downstream node f2'"form", h2 'read from the complementary dictionary data 431A Read "computer", and the node _{y 1.} Since the surplus key is “Keitai”, there is a surplus match with the node f2 ′ “Keitai”. Therefore, the conversion candidate extraction unit 130, a node y ₁ to excess _matching, that is, 'downstream node f3 one of the "form"'"wo" node f2, the complementary dictionary data 431A node g3 'Read "to" with the node y ₂ reads, excess key is updated to delete a node from the "form of" y _{1 (node} f2 ') "of".

Since surplus key is "no", the conversion candidate extraction unit 130 reads the full matching nodes y ₂ specifies "no". Then, by the link pointer of the node g3 'This is a specific node y _2, to identify the spelling supplement dictionary data 431 B, node g3 to "no". The conversion candidate extraction unit 130 reads the node g3 “no” + the node g4 “release date” as the same complement unit node string of the node g3 “no”, and the node y that is the upstream node of the read same complement unit node string _1, i.e., the node by the link pointers f2 ', when a specific node in the spelling complementary dictionary data 431 B f2 to "mobile" as the node Y _1, the node node to the same complementary units node series from Y ₁ node Y ₂ column Y To do. The conversion candidate extraction unit 130 associates the spelling of the node f1 (node X) with the spelling of the node f2 + node g3 + node g4 (node row Y) and the number of appearances of the node g4. Is added to the result table 450.

On the other hand, in S515, if the surplus key is determined not to excess matched against nodes _{y n} (S515: NO), it proceeds directly to S514. An example of this is shown in FIG.
Then, if still remains nodes _{y 1} to be compared (S514: NO), it returns to S509. If the comparison has been completed for all the nodes y ₁ (S514: YES), the process proceeds to S506.

As shown in FIG. 15 (d), when the search key is an unconfirmed character “because the key was broken”, n = 2, that is, the surplus key “because the key was broken” and the node y ₁ “ And, the surplus key “because it was broken” and the node y ₂ “broken” are redundantly matched, but if n = 3, that is, if the surplus key becomes “so”, the node y ₃ cannot be acquired in S510. Therefore, it is determined NO in S511 and S515.
Further, when the search key is the unconfirmed character “Keitai sieving”, when n = 1, that is, when the surplus key “ga sift” and the node y ₁ “ga” are redundantly matched, but n = 2. , that is, if the surplus key becomes "old", because it does not match any node _{y n,} is determined as nO in S511 and S515.

Next, the “first determination process” in S600 in the candidate extraction process shown in FIG. 10 will be described with reference to FIGS. FIG. 16 shows a flowchart of the head determination process executed by the conversion candidate extraction unit 130 in S600 of FIG. 10, and FIG. 17 shows an example of the head determination process.
The search key processed in the flow of FIG. 16 includes a case of “determined part only” and a case of “determined part + unconfirmed part”. The conversion candidate extraction unit 130 acquires one node X that is one downstream of root from the spelling complement dictionary (S601). The definition of the inclusion relationship between the node X and the search key is the same as in the case of the “head unconfirmed process”.
Search key ⊂ Node X Partial match Search key = Node X Exact match Search key ⊃ Node X Surplus match

  Also, in the head determination process, the match determination is made on the condition of “front match”. FIG. 17A shows an example of partial match, complete match, and surplus match with the search key, taking the case where the node X is “mobile” as an example. In other words, if the search key is the confirmed part “mobile”, the partial match is made. If the search key is the confirmed part “mobile”, the complete match is made.

  If there is no match between the search key and the node X (S602: NO), the conversion candidate extraction unit 130 checks whether there is a node X to be compared with the search key remaining in the spelling completion dictionary (S606). ). If there remains a node X to be compared (S606: NO), the process returns to S601, selects one node that is one downstream of the root that has not been compared, and uses this node as the node X for the subsequent processing. I do.

On the other hand, since the comparison is performed for one node downstream of all the roots, when there is no remaining node to be compared (S606: YES), the head determination process ends.
On the other hand, in the case of partial match (S604: YES), the conversion candidate extraction unit 130 reads the spelling of the same complement unit node string of the node X from the spelling complement dictionary and adds it to the result table 450. At this time, the conversion candidate extraction unit 130 reads the number of appearances added to the last node of the same complement unit node sequence from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling. An example of this is shown in FIG. However, in FIG. 17, the number of appearances written in the result table is omitted.

  As shown in FIG. 17B, for example, when the search key is the confirmed character “mobile” and the node X is the node a1 “mobile” in the spelling complement dictionary data 431B, they partially match. The node a1 “mobile” is assigned the complement unit number “1”, and the node a2 “no”, the node c2 “ga”,... Downstream are all assigned the complement unit number “2”. The unit number is different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, and adds the spelling “mobile” and the number of appearances of the node a1 to the result table 450 in association with each other.

  In addition, the search key is the fixed character “mobile”, and the node X is also the node d1 “mobile phone” in the spelling complement dictionary data 431B. A complement unit number “1” is assigned to the node d1 “mobile phone”, a complement unit number “2” is assigned to the node a2 “no” one downstream, and the complement unit numbers are different. Therefore, the conversion candidate extraction unit 130 sets only the node X as the same complement unit node string, associates the spelling “mobile phone” of the node d1 and the number of appearances, and adds it to the result table 450.

  Further, the search key is the confirmed character “mobile”, and the node X is also the node i1 “mobile” in the spelling complement dictionary data 431B. The node i1 “mobile” is assigned the complementary unit number “1”, and the downstream node i2 “no” and the node i3 “screen” are also assigned the same complementary unit number “1”. Therefore, the conversion candidate extraction unit 130 associates the node i1 + node i2 + node i3 with the same complement unit node string and associates the spelling of the nodes i1, i2, and i3 with the number of appearances of the node i3. To the result table 450.

  If nodes to be compared still remain (S606: NO), the process returns to S601. If the comparison has been completed for all nodes X (S606: YES), the head determination process is terminated.

  On the other hand, in the case of a perfect match (S604: NO, S607: YES), the conversion candidate extraction unit 130 determines from the spelling complement dictionary the node X of the spelling complement dictionary and the same complement unit node one node downstream of the node X. The column spelling is read and combined, and added to the result table 450 (S608). At this time, the conversion candidate extraction unit 130 reads the number of appearances added to the last node in the node sequence from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling. An example of this is shown in FIG.

  As shown in FIG. 17C, for example, when the search key is the confirmed character “new” and the node X is the node f1 “new” in the spelling complement dictionary data 431B, they are completely matched. A node f2 “mobile”, a node h2 “personal computer”,... Downstream of the node f1 is assigned a complement unit number “2”, and further nodes f3 and g3 are assigned a complement unit number “3”. The supplement unit numbers are different. Therefore, the conversion candidate extraction unit 130 (1) associates the spelling of the node f1 (node X) with the spelling of the node f2 (same complementary unit node string) and the number of appearances of the node f2, ( 2) A “new personal computer” in which the spelling of the node f1 (node X) and the node h2 (same complementary unit node string) is combined, and the association of the number of appearances of the node h2, etc. are added to the result table 450.

  In addition, when the search key is the confirmed character “new” and the node X is the node j1 “new” in the spelling complement dictionary data 431B, the search key is completely matched. A complement unit number “2” is assigned to the node j2 “personal computer” that is one downstream of the node j1, and a complement unit number “2” is also assigned to the downstream node j3 “no” and the node j4 “release date”. Has been granted. Therefore, the conversion candidate extraction unit 130 combines the spelling of the same complement unit node sequence consisting of the node j1 (node X) and the node j2 + node j3 + node j4, and the number of appearances of the node j4. Are added to the result table 450 in association with each other.

  Next, the conversion candidate extraction unit 130 checks whether there are any remaining nodes to be compared in the spelling complement dictionary (S606). If there are remaining nodes to be compared (S606: NO), the process returns to S601. On the other hand, if there are no remaining nodes to be compared (S606: YES), the head determination process ends.

  Further, when neither partial coincidence nor complete coincidence is made, redundant coincidence is determined (S607: NO). The conversion candidate extraction unit 130 performs a surplus matching process (S700), which will be described later, and checks whether there are any remaining nodes to be compared in the spelling complement dictionary (S606). If there are remaining nodes to be compared (S606: NO), the process returns to S601. On the other hand, if there are no remaining nodes to be compared (S606: YES), the head determination process ends.

Next, the surplus matching process will be described with reference to FIGS. 18 and 19 show flowcharts of surplus processing executed by the conversion candidate extraction unit 130 in S700 of FIG. 16, and FIGS. 20 to 25 show examples of surplus processing.
18, the conversion candidate extraction unit 130, first, by substituting 1 to n (S701), acquires one node _Y 1 next downstream node of the _{n-1 Y n} from spelling complement dictionary (S702). However, since the node Y ₀ is the node X, the node Y ₁ is a node one downstream of the node X. Examples of node _{Y 1} shown in FIG. 20.

As shown in FIG. 20A, for example, when the search key is “mobile” “no” or “mobile” and the node X is the node a1 “mobile” in the spelling completion dictionary data 431B, the node a2 "no", and the node c2 "ga" is selected as the node _{Y 1.}
Also, if the node X is the node i1 "portable" in the complementation dictionary data 431B spelling, node i2 "no" is selected as the node Y _1.

On the other hand, as shown in FIG. 20B, when the search key is “new“ pa ”and the node X is the node f1“ new ”in the spelling complement dictionary data 431B, one downstream node f2“ mobile ” "and, h2" PC "is selected as the node _{Y 1.}
In addition, the search key is "new" path "", if the node X is a node j1 "new" in the supplement dictionary data 431B spelling, node j2 "PC" is selected as the node Y _1.

Next, when the conversion candidate extraction unit 130 generates a remainder key α _n obtained by deleting the node Y _n−1 from the search key, it is determined whether or not the remainder key α _n is an “unconfirmed part” composed of “Hiragana only”. Judgment is made (S703). Here, FIG. 21A shows an example in which S703 is YES, and FIG. 21B shows an example in which S703 is NO. The FIG. 21 (a), the case where the search key is a "portable" of it "" to be a very key alpha ₁ is undetermined portion "that" in FIG. 21 (b), the search key It indicates that the case of "portable" ga "," very key alpha ₁ is included deterministic part and undetermined portion "of" ga "".

If the remainder α _n is an “undetermined part” consisting of “Hiragana only” (S703: YES), the conversion candidate extraction unit 130 moves from the reading completion dictionary one node Y downstream of the node Y _{n−1. The} node yn corresponding to _n is searched and acquired (S704). Then, the conversion candidate extraction unit 130 determines whether or not the remainder key α _n matches forward and partially matches or completely matches the node y _n (S705).

If the remainder key α _n matches forward with the node y _n and also partially matches or completely matches (S705: YES), the conversion candidate extraction unit 130 searches the node Y ₁ to the node Y _n from the spelling complement dictionary. When the spelling up to the same complement unit node string is acquired as the node string Y, the node X and the node string Y are spell-joined and added to the result table 450 (S706). At this time, the conversion candidate extraction unit 130 reads the appearance count of the last node in the node string Y from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling.

On the other hand, when it is determined in S705 that the remainder key α _n is not included in the node y _n (S705: NO), the conversion candidate extraction unit 130 includes the node y _{n in the} remainder key α _n on the condition that forward matching is used. or, in other words, less key alpha _n determines whether the excess matched against nodes _{y n} (S708). When the remainder key α _n is redundantly matched with the node y _n (S708: YES), 1 is added to the current value of n (S709), and the node y _n-1 is deleted from the remainder key α _n-1. When the remainder key α _n is generated (S710), the processing from S704 is performed.
For example, in S709, when the value of n is updated to 2, in S710, the conversion candidate extracting unit 130 generates a key alpha ₂ remainder by remainder obtained by deleting the node _{y 1} from a remainder key alpha _1, in S704, obtaining one downstream node _{y 2} nodes _{y 1.} Next, it is determined whether or not the generated remainder key α ₂ is included in the node y ₂ on the condition that a forward match is made.
The above example is shown in FIG. However, in FIG. 22, the number of appearances written in the result table is omitted. After the process of S706, the conversion candidate extraction unit 130 proceeds to S707.

As shown in FIG. 22 (a), the search key "new" path "," very key alpha ₁ is "Pas", the node X to the node f1 "new", the node Y ₁ node f2 "mobile" node h2 If it is "PC", the conversion candidate extraction unit 130, a node y _1, read supplement dictionary data 431A from node f2 'reading' form ", node h2 'to obtain the reading" PC "in the. Remainder 'for reading to forward matching and broad, the conversion candidate extraction unit 130, the spelling supplement dictionary data 431 B, node h2' key alpha ₁ and node h2 identifying a node h2 corresponding to. The node h2 “PC” is assigned the complement unit number “2”, the node h3 one downstream of the node h2 is assigned the complement unit number “3”, and the complement unit number is different. The candidate extraction unit 130 sets only the node h2 as the same complement unit node string, and sets the same complement unit node string of the node h2 as the node string Y. The node conversion candidate extraction unit 130 associates the spelling of the node f1 (node X) with the spelling of the node h2 (node string Y) and the number of appearances of the node h2, and adds them to the result table 450. .

Also, the search key "new" path "," very key alpha ₁ is "Pas", the node X to the node j1 "new", if the node Y ₁ is the node j2 "PC", the conversion candidate extraction unit 130, as a node _{y 1,} to get the "PC" from reading complement dictionary data 431A node j2 'reading of. The readings of the remainder key α ₁ “Pa” and the node j2 ′ are forward coincident and partial coincidence. The conversion candidate extraction unit 130 reads node j2 “personal computer” + node j3 “no” + node j4 “release date” from the spelling completion dictionary data 431B as the same complement unit node string of the node j2, and sets it as the node string Y. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node j2 + node j3 + node j4 (node string Y) and the number of appearances of the node j4. It adds to the result table 450.

On the other hand, as shown in FIG. 22 (b), "new" PC "," search key, much the key alpha ₁ is "PC", the node X to the node f1 "new", the node Y ₁ node f2 "Mobile "If a node h2" PC ", the conversion candidate extraction unit 130, a node y _1, read supplement dictionary data 431A from node f2 'reading'form", node h2 'to obtain the reading "PC" in . Since the remainder key α ₁ is redundantly coincident with the reading of the node h 2 ′, the conversion candidate extraction unit 130 generates a remainder key α ₂ “no” obtained by deleting the node h 2 ′ (node y ₁ ) from the remainder key α _1. , and node y ₂ reads out the "no"'one downstream node h3 of' node h2 from reading complement dictionary data 431A. Remainder key alpha ₂ and the node h3 'for the exact match, the conversion candidate extraction unit 130, the spelling supplement dictionary data 431 B, as the same complementary units node sequence of nodes h3, node h3 "of" + node h4 "Date" Is specified as the node string Y from the node Y ₁ to the same complementary unit node string of the node Y ₂ . The conversion candidate extracting unit 130 associates the spelling of the node f1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.

Further, when the search key is “new“ pason ”, the remainder key α ₁ is“ pason ”, and the node X is“ new ”, the node j1 (node X) and the node j2 + The “new PC release date” combined with the spelling of node j3 + node j4 (node row Y) and the number of appearances of node j4 are added to result table 450.

On the other hand, when the surplus key α _n does not coincide with the node y _n (S708: NO), that is, in the case shown in FIG. 22C, the process directly proceeds to S707.
As shown in FIG. 22 (c), for example, the search key "portable" from "," less if the key alpha ₁ is "from", be any node X to the node a1, i1 of "portable" , The node y ₁ and the remainder key do not have partial match, complete match or surplus match.
Also, if the search key is “mobile phone“ no mangamen ””, up to n = 2, that is, the remainder key α ₁ “no anger has gone” and node y ₁ “no”, the remainder key α ₂ “Gamen is clear” and node y ₂ “Gamen” are redundantly matched, but when n = 3, that is, when the remainder key α ₃ becomes “clear”, node Y _{4 in} S704 , can not obtain a node _{y 4} is determined as NO in S705 and S 708.

Thereafter, any remaining still Node _{Y 1} to be compared (S707: NO), returns to S701, performs still one node _{Y 1} that is not performed compared selected and subsequent processing. On the other hand, if the comparison of all the nodes _{Y 1} has been completed (S707: YES), ends the surplus matching.

If there is a confirmed character string in the remainder key α _n or only the unconfirmed part is not only hiragana (S703: NO), the conversion candidate extraction unit 130 determines that the remainder key α _n is the node Y _It is determined whether or not _n matches partly or completely (S711). If the remainder key α _n partially matches or completely matches the node Y _n (S711: YES), the conversion candidate extraction unit 130 from the spelling complement dictionary to the same complement unit node string of the node Y _n from the node Y ₁ Is acquired as the node string Y, the spelling of the node X and the node string Y is combined and added to the result table 450 (S712). This example is shown in FIGS. 23 (a) and (b). However, in FIG. 23, the number of appearances written in the result table is omitted. Thereafter, the process proceeds to S707.

In FIG. 23 (a), the search key "mobile", a remainder key alpha ₁ is "no", the node X to the node a1 "portable", when the node Y ₁ is the node a2 "of" conversion candidate The extraction unit 130 determines that the node Y ₁ and the remainder key α ₁ completely match (S711: YES). The conversion candidate extraction unit 130 sets the node a2 “no” + the node a3 “screen” and the node a2 “no” + the node b3 “whether” from the spelling complement dictionary data 431B as the same complement unit node sequence of the node a2. Read, node node Y. The conversion candidate extraction unit 130 associates (1) the “portable screen” in which the spelling of the node a1 and the node a2 + node a3 (node row Y) is combined with the number of appearances of the node a3, and (2) the node a1. Are added to the result table 450, and the association of “the presence of mobile phone” that combines the spelling of the node a 2 + node b 3 (node row Y) and the number of appearances of the node b 3 is added.
Incidentally, also when the node X is the node i1 "mobile", the node Y ₁ node i2 "no", "mobile screen", and the association of the number of occurrences of the node i3 is the result added to the table 450 Is done.

In FIG. 23B, when the search key is “new pa”, the remainder key α ₁ is “pa”, the node X is the node f1 “new”, and the node Y ₁ is the node h2 “personal computer”, the conversion candidate extraction unit 130 determines that the node _{Y 1} and less key alpha ₁ match part (S711: YES). The node h2 “PC” is assigned the complement unit number “2”, the node h3 one downstream of the node h2 is assigned the complement unit number “3”, and the complement unit number is different. The candidate extraction unit 130 sets only the node h2 as the same complement unit node string, and sets the same complement unit node string of the node h2 as the node string Y. The node conversion candidate extraction unit 130 associates the spelling of the node f1 (node X) with the spelling of the node h2 (node string Y) and the number of appearances of the node h2, and adds them to the result table 450. .

Also, if the search key is a "new path", too key alpha ₁ is "Pas", the node X to the node j1 "new", the node Y ₁ is a node j2 "PC", the conversion candidate extraction unit 130, the node Y ₁ and less key alpha ₁ determines that partially match (S711: YES). The conversion candidate extraction unit 130 reads node j2 “personal computer” + node j3 “no” + node j4 “release date” as the same complement unit node string of node j2, and sets it as node string Y. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node j2 + node j3 + node j4 (node string Y) and the number of appearances of the node j4. It adds to the result table 450.

Then, if still remains nodes _{Y 1} to be compared (S707: NO), it returns to S701. On the other hand, if the comparison of all the nodes _{Y 1} has been completed (S707: YES), ends the surplus matching.

On the other hand, when the remainder key α _n does not partially match or completely match the node Y _n (S711: NO), in FIG. 19, the conversion candidate extraction unit 130 determines whether or not the remainder key α _n includes the node Y _n . Determination is made (S713). When the remainder key α _n does not include the node Y _n (S713: NO), that is, in the case as shown in FIG. 23C, the process proceeds to S707 in FIG.
In FIG. 23 (c), the example, "from the mobile" search key, when the remainder key alpha ₁ is "from", be any node X to the node a1, i1 of "mobile", and the node Y ₁ The remainder key α does not match partly or completely, and the remainder key α ₁ does not include the node Y ₁ .

On the other hand, in FIG. 19, when the remainder key alpha _n encompasses the node _{Y n} (S713: YES), the conversion candidate extracting unit 130 further creates a very key alpha _{n + 1,} except for the node _{Y n} from a remainder key alpha _n Then, it is determined whether or not the remainder key α _{n + 1} is an “undetermined part” composed of “Hiragana only” (S714).
Less if the key alpha _{n + 1} is an undetermined portion consisting of only Hiragana (S714: YES), the conversion candidate extraction unit 130, from one downstream node _{Y n + 1} and complementary dictionary read node _{y n + 1} corresponding node _{Y n} When acquired (S715), it is determined whether the remainder key α _{n + 1} partially matches or completely matches the node y _{n + 1} (S716). If the remainder key α _{n + 1} partially matches or completely matches the node y _{n + 1} (S716: YES), the conversion candidate extraction unit 130 determines from the spelling complement dictionary to the same complement unit node string of the node Y ₁ to the node Y _{n + 1.} Is acquired as the node string Y, the spelling of the node X + the node string Y is combined and added to the result table 450 (S717). At this time, the conversion candidate extraction unit 130 reads the appearance count added to the last node of the node string Y from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling.

On the other hand, in S716, when it is determined that the remainder key α _{n + 1} does not partially match or completely match the node y _{n + 1} (S716: NO), the conversion candidate extraction unit 130 uses the remainder key α as a condition that the node y _{n + 1} is a forward match. _It is determined whether it is included in _{n + 1} , that is, whether the remainder key α _{n + 1} is redundantly matched with the node y _{n + 1} (S718). Remainder when the key alpha _{n + 1} is redundant match for the node _{y n + 1 (S718: YES} ), 1 is added to the value of the current n (S719), the key alpha _{n + 1} remainder remove the node _{y n} from the remainder key alpha _n Is generated (S720), the processing from S715 is performed.
For example, in S719, when the value of n is updated to 2, in S720, the conversion candidate extracting unit 130 generates a key alpha ₃ remainder by remainder obtained by deleting the node _{y 2} from a remainder key alpha _2, in S715, obtaining one downstream node _{y 3} nodes _{y 2.} Next, it is determined whether or not the remainder key α ₃ generated in S716 partially matches or forward matches the node y ₃ .

  The above examples are shown in FIGS. 24 (a), (b), and (c). However, in FIG. 24, the number of appearances written in the result table is omitted. After the process of S717, the process proceeds to S707 in FIG.

In FIG. 24 (a), a search key is "portable" ga "", a remainder key alpha ₁ is "a" but "," node X to the node a1 "mobile", the node Y ₁ is a node a2 "of" , The remainder key α ₁ contains the node Y ₁ . Thus, the remainder create a key α ₂ "ga". Less for the key alpha ₂ is undetermined portion consisting of only Hiragana (S714: YES), the conversion candidate extraction unit 130, the node _{Y 2} is the next node of the node _{Y 1,} i.e. corresponding to the node a3 "screen" node a3 'read the "screen" as a node _{y 2} reading of the reading from the complementary dictionary data 431A (S715). Node and _{y 2} at a node a3 '"screen", a remainder key alpha ₂ "ga" is broad match (S716: YES). In spelling supplement dictionary data 431 B, since there is no downstream nodes in the node a3, conversion candidate extracting unit 130, a node a3 only the same complement unit node sequence of node Y _2, the same complement of the node Y ₂ from the node Y ₁ A node string Y is defined up to the unit node string. The conversion candidate extraction unit 130 associates the spelling of the node a1 (node X) with the spelling of the node a2 + node a3 (node row Y) and the number of appearances of the node a3 in the result table 450. to add.
Incidentally, linked by similar processes when the node X to the node i1 "mobile", the node _{Y 1} is the node i2 "no", the node i1 and (node X), node i2 + node i3 spelling (node sequence Y) The association between the “portable screen” and the number of appearances of the node i3 is added to the result table 450.

On the other hand, in FIG. 24 (b), the search key "new computer" of "", a remainder key alpha ₁ is "PC" no "," node X node f1 "new", the node Y ₁ node h2 " In the case of “PC”, the remainder key α ₁ includes the node Y ₁ . Therefore, the remainder key α ₂ “NO” is created. Less for the key alpha ₂ is undetermined portion consisting of only Hiragana (S714: YES), the conversion candidate extraction unit 130, the node _{Y 2} is the next node of the node _{Y 1,} corresponding to the words node h3 "of" node h3 'of reading read from the complementary dictionary data 431A to read the the node _{y 2} "to" (S715). Node and _{y 2} is the node h3 '' of "a remainder key alpha _2" The "exact match to order (S716: YES), the conversion candidate extraction unit 130, as the same complementary units node sequence of nodes h3, node h3 "the" + node and h4 to identify the "Release Date", and node Y ₁ from the node Y ₂ of the same complement unit node column until the node column Y. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.
The node X is the node j1 "new", the same process is when the node _{Y 1} is the node j2 "PC", spelled nodes j1 and (node X), a node j2 + node j3 + node j4 (node sequence Y) Are associated with the “release date of a new personal computer” and the number of appearances of the node j4.

On the other hand, in FIG. 24C, the search key is “new personal computer“ no-ha ””, the remainder key α ₁ is “personal computer“ no-ha ””, the node X is the node f 1 “new”, and the node Y ₁ is the node In the case of h2 “personal computer”, the remainder key α ₁ includes the node Y ₁ . Therefore, the remainder key α ₂ “noha” is created. Less for the key alpha ₂ is undetermined portion consisting of only Hiragana (S714: YES), the conversion candidate extraction unit 130, the node _{Y 2} is the next node of the node _{Y 1,} corresponding to the words node h3 "of" node h3 'of reading read from the complementary dictionary data 431A to read the the node _{y 2} "to" (S715).
Since the node y ₂ and the remainder key α ₂ are redundantly matched (S716: NO, S718: YES), the conversion candidate extraction unit 130 deletes the node y ₂ from the remainder key α ₂ “noha”, and the remainder key α ₃ “Ha” is created (S719, S720). The conversion candidate extraction unit 130 reads from the supplementary dictionary data 431A the reading “hatsubibi” of the node h4 ′ that is one downstream of the node h3 ′ (node y ₂ ) as the node y ₃ (S715).

The node y ₃ and the remainder key α ₃ partially match (S716: YES). In spelling supplement dictionary data 431 B, since there is no downstream nodes in the node h4, the conversion candidate extraction unit 130, the node h4 only the same complement unit node sequence of the node Y _3, the same complement of the node Y ₃ from the node Y ₁ A node string Y is defined up to the unit node string. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.
The node X is the node j1 "new", the same process is when the node _{Y 1} is the node j2 "PC", spelled nodes j1 and (node X), a node j2 + node j3 + node j4 (node sequence Y) Are associated with the “release date of a new personal computer” and the number of appearances of the node j4.

On the other hand, if the surplus key α _{n + 1} does not coincide with the node y _{n + 1} (S718: NO), that is, as shown in FIG. 24D, the process proceeds to S707 in FIG.

For example, as shown in FIG. 24D, the search key is “new personal computer“ GA ””, the remainder key α ₁ is “PC“ GA ””, node X is node f 1 “new”, and node Y ₁ is In the case of the node h2 “personal computer”, the remainder key α ₂ “ga” is created. The reading “NO” of the node Y ₂ , that is, the node h 3 ′ (node y ₂ ) corresponding to the node h 3 “NO” has no partial match, complete match, or redundant match with the remainder key α ₂ . Node X node j1 "new", node Y ₁ is also the same in cases where it is a node j2 "PC".
Then, if still remains downstream node _{Y 1} to be compared (S707: NO), it returns to S701. On the other hand, if the comparison of all the nodes _{Y 1} has been completed (S707: YES), ends the surplus matching.

On the other hand, in FIG. 19, when the remainder key α _{n + 1} is not an “undecided part” consisting of “Hiragana only” (S714: NO), the conversion candidate extraction unit 130 has the remainder key α _{n +} 1 one downstream of the node Y _n. It is determined whether the node Yn _{+ 1} partially matches or completely matches (S721). Less if the key alpha _{n + 1} and the node _{Y n + 1} are partial or exact match (S721: YES), the conversion candidate extraction unit 130, spelling than spelling complement dictionary, from the node _{Y 1} to node _{Y n + 1} of the same complementary units node sequence Is acquired as the node string Y, the node X + the node string Y is added to the result table 450 (S722). At this time, the conversion candidate extraction unit 130 reads the appearance count of the last node in the node string Y from the spelling complement dictionary, and writes it in the result table 450 in association with the spelling.

On the other hand, when it is determined in S721 that the remainder key α _{n + 1} is not included in the node y _{n + 1} (S721: NO), the conversion candidate extraction unit 130 includes the node y _{n + 1 in the} remainder key α _{n + 1} on the condition of a forward match. That is, it is determined whether the remainder key α _{n + 1} is redundantly matched with the node y _{n + 1} (S723). If the remainder key α _{n + 1} is redundantly matched with the node y _{n + 1} (S723: YES), 1 is added to the current value of n (S724), and the processing from S714 is performed.
For example, in S724, when the value of n is updated to 2, in S714, the conversion candidate extracting unit 130 generates a key alpha ₃ remainder by remainder obtained by deleting the node _{Y 2} from the remainder key alpha _2, remainder key alpha ₃ determines whether the hiragana is unconfirmed.
The above examples are shown in FIGS. 25 (a), (b), and (c). However, in FIG. 25, the number of appearances written in the result table is omitted. After the process of S722 or when it is determined in S723 that the surplus key α _{n + 1} does not coincide with the node y _{n + 1} (S723: NO), the conversion candidate extraction unit 130 proceeds to S707 in FIG.

In FIG. 25 (a), a example, the search key is "new PC", a remainder key alpha ₁ is "PC", the node X to the node f1 "new", the node Y ₁ node h2 is "PC" In this case, since the remainder key α ₁ “personal computer” includes the node h 2 “personal computer” that is the node Y ₁ , the remainder key α ₂ “no” is created. Since the node Y ₂ , that is, the node h 3 “no” and the remainder key α ₂ completely match (S 721: YES), the conversion candidate extraction unit 130 sets the node h 3 “no” + as the same complement unit node string of the node h 3. When you identify the node h4 "Release Date", from the node Y ₁ to the same complement unit node column of the node Y ₂ and node row Y. The conversion candidate extracting unit 130 associates the spelling of the node f1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.

The node X is the node j1 "new", the same process is when the node _{Y 1} is the node j2 "PC", spelled nodes j1 and (node X), a node j2 + node j3 + node j4 (node sequence Y) Are added to the result table 450. The “new PC release date” and the number of appearances of the node j4 are associated with each other.

Figure 25 (b) as shown in, for example, the search key "calling of new personal computer" is a very key α ₁ is "the calling of the personal computer", the node X is node f1 "new", node Y ₁ is node h2 In the case of “personal computer”, since the remainder key α ₁ “departure of personal computer” includes the node h2 “personal computer” which is node Y ₁ , the conversion candidate extraction unit 130 creates the remainder key α ₂ “departure”. to together, read from spelling supplement dictionary data 431B node h2 spelling h3 is one downstream node of the "PC", "no" as a node Y _2.
Remainder key alpha ₂ Node _{Y 2} and excess matched for (S721: NO, S723: YES ), the conversion candidate extraction unit 130, too key alpha ₂ from the node h3 (node _{Y 2)} less key alpha ₃ deleting the " Generate ". Since much key alpha ₃ is not undetermined hiragana (S714: NO), the conversion candidate extraction unit 130, spelling node h3 "of" is one downstream node of h4 spelling the "Date" as the node Y ₃ Read from the complementary dictionary data 431B.

Remainder key alpha ₃ coincides node _{Y 3} and part (S721: YES). In spelling supplement dictionary data 431 B, since there is no downstream nodes in the node h4, the conversion candidate extraction unit 130, the node h4 only the same complement unit node sequence of the node Y _3, the same complement of the node Y ₃ from the node Y ₁ A node string Y is defined up to the unit node string. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.

The node X is the node j1 "new", the same process is when the node _{Y 1} is the node j2 "PC", spelled nodes j1 and (node X), a node j2 + node j3 + node j4 (node sequence Y) Are associated with the “release date of a new personal computer” and the number of appearances of the node j4.

As shown in FIG. 25C, for example, the search key is “new PC“ ha ””, the remainder key α ₁ is “PC“ ha ””, node X is node f1 “new”, node Y _{When 1} is the node h2 “personal computer”, the remainder key α ₁ “the personal computer“ ha ”” includes the node h2 “personal computer” that is the node Y ₁ , so the conversion candidate extraction unit 130 uses the remainder key α ₂ "a" is "" with creating a read from spelling supplement dictionary data 431B node h2 spelling h3 is one downstream node of the "PC", "no" as a node Y _2.
Remainder key alpha ₂ Node _{Y 2} and excess matched for (S721: NO, S723: YES ), the conversion candidate extraction unit 130, too key alpha ₂ from the node h3 (node _{Y 2)} less key alpha ₃ deleting the " Is generated. Since the remainder is key alpha ₃ is undetermined hiragana (S714: YES), the conversion candidate extraction unit 130, node h3 (node _{Y 2)} node h3 corresponding to '(node _{y 2)} 1 next downstream node of the h4' of reading the "sale beauty", read from the complementary dictionary data 431A read as node _{y 3} (S715).

The node y ₃ and the remainder key α ₃ partially match (S716: YES). In spelling supplement dictionary data 431 B, since there is no downstream nodes in the node h4, the conversion candidate extraction unit 130, the node h4 only the same complement unit node sequence of the node Y _3, the same complement of the node Y ₃ from the node Y ₁ A node string Y is defined up to the unit node string. The conversion candidate extraction unit 130 associates the spelling of the node j1 (node X) with the spelling of the node h2 + node h3 + node h4 (node row Y) and the number of appearances of the node h4. It adds to the result table 450.
The node X is the node j1 "new", the same process is when the node _{Y 1} is the node j2 "PC", spelled nodes j1 and (node X), a node j2 + node j3 + node j4 (node sequence Y) Are associated with the “release date of a new personal computer” and the number of appearances of the node j4.

After the process of S722 or when it is determined in S723 that the surplus key α _{n + 1} does not coincide with the node y _{n + 1} (S723: NO), the process proceeds to S707 in FIG. FIG. 25 (d) shows an example of a case where there is no partial match, complete match or surplus match.

In FIG. 25D, for example, the search key is “price” of the new computer, the remainder key α ₁ is “price” of the computer, the node X is “new” and the node Y ₁ is the node In the case of h2 “personal computer”, the surplus key α ₂ “no price” and the node h3 “no” of the node Y ₂ have a coincidence, but the surplus key α ₃ “price” and one of the nodes h3 There is no partial match, complete match or surplus match with downstream nodes. Node X node j1 "new", node Y ₁ is also the same in cases where it is a node j2 "PC".

Thereafter, 18, any remaining still downstream node _{Y 1} to be compared (S707: NO), returns to S701. On the other hand, if the comparison of all the nodes _{Y 1} has been completed (S707: YES), ends the surplus matching.

Next, the output process in S400 of FIG. 7 will be described. FIG. 26 shows a flowchart of output processing executed by the conversion candidate output unit 140.
In the figure, the conversion candidate output unit 140 leaves only one of these overlapping character strings when the character string written in the result table 450 includes the same character string ( S401). For ease of explanation, the processing from S402 to S405 will be described separately when the search key is “tail part only”, “tail part + unconfirmed part”, and “unconfirmed part only”.

<When the search key is the last part only>
The conversion candidate output unit 140 deletes the “tail part” of the search key from the character string in the result table 450 (S402). Next, it is determined whether there is an output candidate word that has been completely matched with the “undefined part” of the search key as a result of deletion (S403). In this example, the search key does not include “indeterminate part”, so S403 is automatically NO. As a result, an output as shown in FIG. 27A is obtained (S405).

  In FIG. 27A, the search key is “new”, and the character strings “new mobile phone”, “new personal computer”, and “new personal computer release date” are stored in the result table. At this time, “mobile”, “personal computer”, and “personal computer release date” excluding “new” from each character string stored in the result table are output as output candidate words in descending order of appearance frequency. A screen display example of the output candidate words thus shaped is shown in FIG.

<When the search key is "end part + unconfirmed part">
First, the conversion candidate output unit 140 deletes the “tail part” of the search key from the character string of the result table 450 (S402). Furthermore, if the output candidate word includes an output candidate word that completely matches the undetermined part of the search key as a result of the deletion of the tail part (S403: YES), the matching output candidate word is selected as the output candidate word of the result table 450. (S404).

  In example 1 shown in FIG. 27B, the search key is “new“ pa ”consisting of the suffix“ new ”+ the unconfirmed part“ pa ”, and the result table includes the strings“ new PC ”,“ new ” "Release date of PC" is stored. In the node column “PC” and “PC release date” of the output candidate word after deleting the tail part, there is no node that exactly matches the undetermined part “pa” of the search key, so “PC” and “PC release date” Are output as output candidate words in the order of appearance frequency (S405). A screen display example of the output candidate words thus shaped is shown in FIG.

  In the example 2 shown in FIG. 27C, the search key is “mobile phone” composed of the tail part “mobile phone” + the unconfirmed part “ga”, and the character string “mobile phone is broken” is displayed in the result table. It is remembered. The unconfirmed part “GA” of the search key matches “GA” in the node string “GA / broken” of the output candidate word after the end part is deleted. However, since it is not a perfect match (S403: NO), “is broken” is output as an output candidate word as it is (S405). FIG. 28C shows a screen display example of the output candidate word thus shaped. When the operator selects, for example, “broken”, the unconfirmed “ga” is deleted and “broken” is inserted into the input area with the definite information. When there are a plurality of output candidate words, they are displayed in descending order of the number of appearances.

  Further, in Example 3 shown in FIG. 27D, the search key is “portable“ is there ”” composed of the tail part “mobile” and the unconfirmed part “is there”, and the character string “ "Is there a mobile phone?" Since the unconfirmed part “Are there” of the search key completely matches the output candidate word “Are there” after the end part deletion, the whole “Are?” Is deleted in S404, and the output candidate word is not output in S405. . When the unconfirmed character string already displayed in the input area matches the character string displayed in the output candidate word display area, the operator selects the character in the input area rather than selecting the output candidate word from the output candidate word display area. This is because it is more natural to determine the column using, for example, the Enter key. A screen display example in this case is shown in FIG.

<When the search key is "Only unconfirmed part">
If the search key is “only unconfirmed part”, there is no part deleted in S402, and the search result is output as it is (S405). FIG. 27E shows the case of Example 1, and FIG. 27F shows the case of Example 2. In addition, examples of screen display of output candidate words in such a case are shown in FIGS.

  In Example 1 shown in FIG. 27E, the search key is “Kei”, and the character strings “mobile phone”, “mobile phone”, and “mobile phone screen” are stored in the result table. Accordingly, as shown in FIG. 28 (e), “mobile phone”, “mobile phone”, and “mobile phone screen” are output as output candidate words in descending order of appearance frequency.

  In Example 2 shown in FIG. 27F, the search key is “new”, and the character strings “new mobile phone”, “new personal computer”, and “new personal computer release date” are stored in the result table. Therefore, as shown in FIG. 28 (f), “new personal computer”, “new personal computer release date”, and “new mobile phone” are output as output candidate words in descending order of appearance frequency.

  Exceptionally, if the spelling completion dictionary includes a node consisting only of hiragana, and that node becomes an output candidate word, YES is obtained in S403, and the matched output candidate word is deleted from the result table (S404). . An example of this is shown in FIG.

  After the above S405, the operator operates the keyboard 211 or the like, and selects a suitable candidate from the displayed conversion candidates. When an input signal corresponding to the selection operation is input from the input control device 200 to the control unit 100 (S406: YES), the character string input unit 150 performs confirmation input of the selected conversion candidate (S407).

  Here, the character string input unit 150 determines which of the selection candidates displayed by the conversion candidate output unit 140 has been selected, and controls the output control device 300 to determine the selected conversion candidate as a confirmed character string. Is displayed in the input area. Further, the character string input unit 150 writes the sentence composed of the confirmed character string that has been confirmed and input in this manner, together with the input area identification and time information, in the input sentence storage area 460.

Then, after the confirmation input, the search key process of S200 of FIG. 7 is executed again to generate a search key. Therefore, in the candidate extraction process of S300 in FIG. 7, the head determination process of S600 shown in FIG. 10, that is, the process shown in FIG. 16 is executed.
Alternatively, even after the confirmation input, when the user inputs some unconfirmed character using the keyboard 211, the search key process of S200 of FIG. 7 is executed, and the search key of the tail part + unconfirmed part is generated. Therefore, in the candidate extraction process of S300 in FIG. 7, the head unconfirmed process of S500 shown in FIG. 10, that is, the process shown in FIG. 11 is executed.

  If the conversion candidate is not selected and an input canceling operation is performed due to an input error in the undetermined character string (S406: NO, S408: YES), the process ends without performing the confirmed input. .

  Next, the supplement unit number assigning process will be described. In this complement unit number assigning process, a node sequence to which the same complement unit number should be assigned is detected from the current dictionary data, and a new node sequence generated by assigning the same complement number to the identified node sequence To update the dictionary data. There are two types of node sequences to which the same complement unit number is to be assigned: (1) those that are frequently input to the input area, and (2) those that have few dependency branches. In (1), it is possible to select an expression that is frequently input by the operator as a conversion candidate. In (2), when a certain word is used only in some limited expressions, the expression is displayed as a conversion candidate.

A complementary unit number assigning process when assigning the same complementary unit number to a node string that is frequently input to the input area will be described with reference to FIGS.
First, it is assumed that a pair of reading completion dictionary data and spelling completion dictionary data corresponding to each input area is stored in the input content dictionary area 430 in advance.

29 shows a flowchart of the complementary unit number assigning process executed by the dictionary generation unit 160, FIG. 30 shows an example of data of the complementary unit number assigning process of FIG. 29, and FIG. 31 shows an example of updating dictionary data. .
In FIG. 29, the dictionary generation unit 160 selects a supplementary unit number assignment process from the input sentence storage area 460 by a predetermined period such as one week, every month, or when a user inputs an instruction. The input sentence in the input area corresponding to the dictionary data to be read is read out. At this time, all the input texts in the input area may be read, or texts for a predetermined period such as three months before the present time may be read (S801). Note that information for specifying dictionary data to be subjected to the complementary unit number assigning process may be input in advance, or may be sequentially performed for all dictionary data. The dictionary generation unit 160 divides the text read in S801, that is, the character string in accordance with the sentence end symbol, and generates a sentence list (S802). The sentence ending symbol can be, for example, a punctuation mark “.”.

The dictionary generation unit 160 acquires sentences that have not yet been read in order from the top of the sentence list generated in S802 (S803). The dictionary generation unit 160 deletes unnecessary characters, for example, blanks, tabs, line feeds, and the like other than those sandwiched between alphanumeric characters from the sentence read in S803 (S804). The dictionary generation unit 160 performs syntax analysis of the sentence from which unnecessary characters are deleted in S804 (S805). For this parsing process, any conventional parsing software that extracts words from the input sentence or the dependency relationship between words can be used.
When analysis fails (S806: NO), it progresses to S803.

  When the analysis is successful (S806: YES), the dictionary generation unit 160 generates a subtree array list after extracting the parent-child relation subtree and the single subtree from the analysis result (S807). An example of the subtree array list is shown in FIG. The subtree array list indicates, for each subtree, its own subtree number, the reading and spelling of words constituting the subtree, the spelling and part of speech, and the number of subtrees in a parent-child (dependency) relationship. Of the subtrees arranged vertically, the subtree located at the top is the parent, and the subtree located at the bottom is the child.

  The dictionary generation unit 160 writes the path (matching path) corresponding to the spelling and reading obtained from the subtree to the path file (S808). An example of this is shown in FIG. In other words, a path is generated as a dependency-related clause with the child of the parent-child relationship subtree as the dependency side and the parent as the reception side, and further, the words that make up each subtree (sentence) are read and spelled as nodes. Generate a path. Then, the particle at the last node of the child-side subtree in the parent-child relationship, that is, the last node of the dependency-side clause, moves to the parent-side subtree, that is, the first node of the receiving-side clause. The dictionary generation unit 160 assigns a complement unit number in units of phrases. At this time, a complement unit number of 1 is assigned to the head phrase, and a complement unit number of 1 greater than that of the dependency phrase is assigned to the receiver phrase. Then, the complement unit number assigned to each clause is assigned to the node constituting the clause.

  If there is an unprocessed sentence in the sentence list (S809: YES), the dictionary generation unit 160 performs the process from S602 again. When processing has been completed for all sentences included in the sentence list (S809: NO), the dictionary generation unit 160 stores the intermediate file composed of the generated path file in the storage area held by itself (S810).

  The dictionary generation unit 160 acquires all the node strings having two or more consecutive complementary unit numbers from the intermediate file stored in S810. For example, when the reading complement dictionary data and the spelling complement dictionary data are composed of three or less phrases, the complementing unit number is a continuous node string such as 1-2, 2-3, 3-4,. Read consecutive node strings such as 2-3, 2-3-4, 3-4-5,. Then, when the number of appearances is counted for each node sequence composed of the same spelling, the appearance frequency of each node sequence, for example, the standard deviation of the number of appearances, and the appearance ratio in the whole are obtained (S811). The appearance frequency itself may be used as the appearance frequency. Then, it is determined that a node string having an appearance frequency higher than a predetermined threshold is a node string to be assigned the same complement unit number (hereinafter referred to as “target node string”). When the dictionary generation unit 160 specifies the target node string from the spelling complement dictionary data, the node string including the same node as the target node string and the node string including the target node string are one of the root nodes of the spelling complement dictionary data. The same complement unit number is assigned to the node sequence that matches the target node sequence for the added node sequence. Then, the node sequence corresponding to the added node sequence is read and added to the complementary dictionary data (S812). An example of this is shown in FIG.

For example, it is assumed that the response input dictionary data (A1) 431 is a processing target and the target node string is “mobile / buy / buy”. In this case, the dictionary generation unit 160 identifies the node f2 + node f3 + node f4 from the spelling complement dictionary data 431B (FIG. 5). As shown in FIG. 31A, the dictionary generation unit 160 includes a new node s1 “mobile” -node s2 “buy” -node s3 “buy” and a new node t1 “new”- A node string consisting of node t2 “mobile” -node t3 “has” -node t4 “buyed” is added under the root of the spelling complement dictionary data 431B. Then, the complement unit number “1” is assigned to the nodes s1, s2, and s3. Further, the complement unit number “1” is assigned to the node t1, and the complement unit number “2” is assigned to the nodes t2, t3, and t4.
Further, as shown in FIG. 31B, the dictionary generation unit 160 adds the new node s1 ′ “Keitai” -node s2 ′ “to” -node s3 ′ “Kai” to the reading completion dictionary data 431A. And a new node t1 ′ “new” -node t2 “Keitai” -node t3 “is” -node t4 “Kato” is added under the root. Then, the complement unit number “1” is assigned to the nodes s1 ′, s2 ′, and s3 ′. Further, the complement unit number “1” is assigned to the node t1 ′, and the complement unit number “2” is assigned to the nodes t2 ′, t3 ′, and t4 ′.

  Further, in S812, the dictionary generation unit 160 further has the same complement unit number in the spelling complement dictionary data and the reading complement dictionary data in the node sequence whose appearance frequency is lower than a predetermined threshold and whose complement units are continuous. If there is a node string assigned and registered, the node string under the root including this node string is deleted. For example, it is assumed that the response input dictionary data (A1) 431 is a processing target, and the appearance frequency of “PC /// release date” is equal to or less than a threshold value. The dictionary generation unit 160 identifies the node j2 + node j3 + node j4 to which the same complement unit number “2” is assigned from the spelling complement dictionary data 431B (FIG. 5). In this case, the node strings of the nodes j1 to j4 are deleted from the spelling complement dictionary data 431B, and the node strings of the nodes j1 'to j4' are deleted from the reading complement dictionary data 431A.

  Further, the dictionary generation unit 160 updates the number of appearances described in each node of the reading completion dictionary data and the spelling completion dictionary data with the number of appearances counted for each node sequence.

  Next, a complementary unit number assigning process when assigning the same complementary unit number to those with few dependency branches will be described with reference to FIGS. 32 and 33. 32 shows a flowchart of the complementary unit number assigning process executed by the dictionary generation unit 160, and FIG. 33 shows an example of dictionary update in the complementary unit number assigning process of FIG.

  In FIG. 32, the dictionary generation unit 160 acquires one upstream node that is one node downstream from the root from the spelling complement dictionary data of the dictionary data that is the target of the supplement unit number assignment process. This is designated as node X (S901). The dictionary generation unit 160 adds up the number of nodes one downstream of the node X, the number of one downstream node of each downstream node,..., And calculates the number of node strings starting with the node X. If acquired (S902), it is determined whether or not the number of acquired node strings is equal to or less than a predetermined value (S903).

  When the number of node strings is equal to or less than the predetermined number (S903: YES), the dictionary generation unit 160 further determines whether the complementary unit numbers of the nodes downstream of the node X are the same (S904). If they are not the same (S904: NO), the dictionary generation unit 160 duplicates all the node strings having the node X as the most upstream, assigns the same complement unit number “1”, and adds them to the spelling complement dictionary data. Then, the node sequence corresponding to the added node sequence is read and added to the complementary dictionary data (S905). This example is shown in FIG.

As shown in FIG. 33A, the most upstream node of the spelling completion dictionary data is the node a1 “mobile”, the node one node downstream of the node a1 is the node a2 “no” and the node c3 “is”, the node One node downstream of a2 “no” is node a3 “screen” and node b3 “is there”, one node downstream of node c3 “ga” is one of node c3 “broken”, “ It is assumed that there are no nodes downstream of “screen”, “is there”, and “broken”. In this case, the dictionary generation unit 160 has 3 node strings starting from the node a1. Since the complement unit numbers of the node a1 and the nodes a2 and c2 are different, the dictionary generation unit 160 duplicates all the node sequences having the node a1 as the most upstream and adds them under the root of the spelling complement dictionary data. A new link pointer is assigned to each node in the node sequence, and a complement unit number “1” is assigned to all nodes a1 and downstream of the node a1.
Furthermore, as shown in FIG. 33 (b), the dictionary generation unit 160 duplicates the node sequence having the node a1 ′ as the most upstream in the reading completion dictionary data and adds it to the root of the reading completion dictionary data. A link pointer corresponding to the reading completion dictionary data and a completion unit number “1” are assigned to each node of the added node row.

When it is determined in S903 that the number of node columns exceeds a predetermined number (S903: NO), when it is determined in S904 that the complementary unit number of each node downstream of the node X is the same (S904: YES), or After performing the processing of S905, the dictionary generation unit 160 checks whether there are any remaining nodes to be compared in the spelling complement dictionary data (S906). If there remains a node to be compared (S906: NO), the process returns to S901, and selects the most upstream node that has not been processed yet, and the subsequent processing is performed with this node as node X.
On the other hand, since all the most upstream nodes have been compared, and there are no remaining nodes to be compared (S906: YES), the head unconfirmed process ends.

  In FIG. 32, the most upstream node that should have the same complement unit number may be detected using the reading complement dictionary data.

  Note that the above complementing unit number assigning process may be performed at the time of dictionary creation. In this case, in S801 of FIG. 29, instead of reading out the input sentence from the input sentence storage area 460, an arbitrary sentence group is read and used.

The embodiment described above is an example, and the scope of application of the present invention is not limited to this. That is, various applications are possible, and all embodiments are included in the scope of the present invention.
For example, in the above embodiment, the dictionary data for each input content is constructed by associating a pair of reading completion dictionary data and spelling completion dictionary data with the input area. What is associated is not limited to the input area. For example, an input screen for inputting a character string or an application to be used may be associated with dictionary data.

  In the above embodiment, the example of the response input dictionary data is mainly shown as the dictionary data according to the input content. However, if it is possible to have a hierarchical structure of words having a dependency relationship with each other, the input content is various contents. Another dictionary data can be constructed. For example, it is possible to build dictionary data used for address input by creating a hierarchical structure of prefecture names and subsequent subordinate regional names, and by creating a hierarchical structure with combinations of surnames and subsequent names. Dictionary data used for personal name input can be constructed. That is, if it is possible to connect a plurality of target words with relative positions, dictionary data according to input contents according to various contents can be constructed.

  Moreover, although the case where a character string was input with the keyboard 211 was illustrated in the said embodiment, as long as a character string can be input, the input method is not restricted to this but is arbitrary. For example, even when a character string is input by voice recognition, more efficient character string input can be realized by applying the present invention.

  In the above embodiment, the case where the present invention is applied to an input terminal of a call center that needs to input the contents of a telephone response in real time has been exemplified. The present invention is particularly suitable for the field where such high-speed input work is required, but is not limited to this, and the present invention can be applied to input devices used in various scenes for inputting character strings. In any case, more efficient character string input can be realized.

  The input terminal 1 according to the above embodiment can be configured not only as a stand-alone but also as a so-called server-client system. That is, for example, as shown in FIG. 34, a client 500 including a control unit 501 and a communication unit 502 is provided with functions of an input device 210 and an output device 310, and the client 500, control unit 601, communication unit 602, storage It is good also as a structure from the server 600 provided with the part 603 and given the search function. The server 600 includes, as a search function, a dictionary selection unit 601-1, a search key generation unit 601-2, a conversion candidate extraction unit 601-3, a conversion candidate output unit 601-4, a character string input unit 601-5, and a dictionary generation unit. 601-6. The dictionary selection unit 601-1, search key generation unit 601-2, conversion candidate extraction unit 601-3, conversion candidate output unit 601-4, character string input unit 601-5, and dictionary generation unit 601-6 are respectively shown in FIG. The same functions as the dictionary selection unit 110, search key generation unit 120, conversion candidate extraction unit 130, conversion candidate output unit 140, character string input unit 150, and dictionary generation unit 160 shown in FIG. The storage unit 603 stores the same data as the storage unit 400 illustrated in FIG.

  Such a program providing method is arbitrary. For example, the program data can be stored and distributed in a storage medium such as a CD-ROM, and the program data can be superposed on a carrier wave so that a predetermined communication medium (for example, It can also be distributed via the Internet).

DESCRIPTION OF SYMBOLS 1 Input terminal 100 Control part 110 Dictionary selection part 120 Search key generation part 130 Conversion candidate extraction part 140 Conversion candidate output part 150 Character string input part 160 Dictionary generation part 200 Input control apparatus 210 Input apparatus 211 Keyboard 212 Pointing device 300 Output control apparatus 310 Output device 311 Display device 400 Storage unit 410 Dictionary storage area 420 Basic dictionary area 430 Dictionary area according to input contents 431 Response input dictionary data (A1)
431A Reading completion dictionary data 431B Spelling completion dictionary data 432 Address input dictionary data (A2)
432A Reading completion dictionary data 432B Spelling completion dictionary data 440 Program storage area 450 Result table 460 Input sentence storage area

Claims (9)

A character string input device that searches and presents conversion candidates according to an input character string and performs character string input by selecting the presented conversion candidate,
A dictionary in which readings before conversion processing of input character strings are recorded as words, and reading completion dictionary data in which words having a dependency relationship are configured as nodes in a tree structure, and the reading completion dictionary A dictionary in which spelling after each word included in the data is converted is recorded as a word, and the tree structure indicated by the reading completion dictionary data by associating the word before conversion processing and the word after conversion processing as a node; A storage unit configured to store the spelling completion dictionary data in which the same complementing unit number is assigned to the nodes in the dependency relationship, which is configured as a matched tree structure and regarded as one node sequence;
A search key generation unit that generates a search key based on user input;
Based on the search key generated by the search key generation unit, the reading completion dictionary data and the spelling complement dictionary data are searched to identify a node of the spelling complement dictionary data corresponding to the search key, and the spelling of the specified node A conversion candidate extraction unit that reads out from the spelling complement dictionary data and combines the spelling of the downstream node of the identified node assigned the same complement unit number as the identified node, and
A conversion candidate output unit that outputs the conversion candidate acquired by the conversion candidate extraction unit;
A character string input device comprising: When the search key is composed of unconfirmed hiragana, the conversion candidate extraction unit detects the most upstream node that is forward-matched and partially matched with the search key from the reading completion dictionary data, and corresponds to the detected node The candidate of the spelling completion dictionary data is identified, and a conversion candidate is obtained by combining the spelling of the identified node and the spelling of the downstream node of the identified node to which the same complement unit number as the identified node is assigned. get,
The character string input device according to claim 1. When the search key is an unconfirmed hiragana, the conversion candidate extraction unit detects the most upstream node that completely matches the search key from the reading completion dictionary data, and the spelling completion dictionary corresponding to the detected node The data node and the one downstream node are identified, the spelling of the identified node and the one downstream node, and the one downstream node assigned with the same complement unit number as the one downstream node To obtain a conversion candidate that combines the further node spelling of
The character string input device according to claim 1. When the search key is a fixed character, the conversion candidate extraction unit specifies the most upstream node that matches the search key forward and partially matches the spelling completion dictionary data, spells the specified node, and specifies the specified To obtain a conversion candidate that combines the spelling of the downstream node of the identified node assigned the same complement unit number as
The character string input device according to claim 1. When the search key is a fixed character, the conversion candidate extraction unit specifies the most upstream node that completely matches the search key from the spelling completion dictionary data and a node that is one downstream of the specified node, and A conversion candidate is obtained by combining the spelling of the identified node and one downstream node and the further downstream node spelling of the one downstream node assigned the same complement unit number as the one downstream node. To
The character string input device according to claim 1. The conversion candidate extraction unit
When the search key is composed of an unconfirmed character string consisting of a confirmed character and an unconfirmed hiragana, by using the confirmed character of the search key, the most upstream node of the spelling completion dictionary data is searched to identify a matching node,
Determining a match between a node downstream of the node of the reading completion dictionary data corresponding to the identified node and an unconfirmed hiragana of the search key;
The identified node from the spelling completion dictionary data, a node downstream of the identified node, and a node corresponding to the node of the reading completion dictionary data determined to match the unconfirmed hiragana of the search key; Read the spelling with a further downstream node assigned the same complement unit number as the downstream node, and obtain a conversion candidate that combines the read spelling.
The character string input device according to claim 1.   A new word generated by identifying a word string having an appearance frequency higher than a predetermined threshold from the input sentence data and assigning the same complement unit number to the node string of the spelling complement dictionary data corresponding to the identified word string 7. The apparatus according to claim 1, further comprising: a dictionary generation unit that adds a node sequence to the spelling completion dictionary data and adds a node sequence corresponding to the added node sequence to the reading completion dictionary data. The character string input device according to any one of the items.   For each uppermost node of the spelling completion dictionary data, the number of node rows starting from the uppermost node is obtained, and when the obtained number of node rows is less than or equal to a predetermined number, A new node string generated by assigning the same complement unit number to the first node string is added to the spelling complement dictionary data, and a node string corresponding to the added node string is added to the reading complement dictionary data The character string input device according to any one of claims 1 to 6, further comprising a dictionary generation unit that performs the operation. A character string input method used for a character string input device that searches for and presents conversion candidates according to an input character string, and inputs a character string by selecting the presented conversion candidate,
The input device is a dictionary in which readings before converting input character strings are recorded as words, and reading completion dictionary data in which words having dependency relations are configured as nodes in a tree structure, and A dictionary in which spelling after each word included in the reading completion dictionary data is converted as a word is recorded as a word, and the word before conversion processing and the word after conversion processing are associated with each other as a node. Comprising a storage unit for storing spelling completion dictionary data in which the same complementing unit number is assigned to a node in a dependency relationship, which is configured as a tree structure that matches the tree structure indicated by
A search key generation process in which the search key generation unit of the character string input device generates a search key based on a user input;
The spelling completion dictionary corresponding to the search key by the conversion candidate extraction unit of the character string input device searching the reading completion dictionary data and the spelling completion dictionary data based on the search key generated in the search key generation process. The data node is identified, and the spelling of the identified node and the spelling of the downstream node of the identified node assigned the same complement unit number as the identified node are read and combined from the spelling complement dictionary data Conversion candidate extraction process for acquiring the converted conversion candidates;
A conversion candidate output process in which the conversion candidate output unit of the character string input device outputs the conversion candidates acquired in the conversion candidate extraction process;
A character string input method characterized by comprising:

Images