JPH1011432A - Character string estimation method and document preparing device using the estimation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate even an unregistered word based on its head several characters in the next and following input steps as long as the word is once inputted by deciding the connection between a character string that is selected among those candidates of estimated character strings and a character string that is stored in a character string buffer in the preceding retrieval step. SOLUTION: A connection decision part 110 refers to the character types or the grammar information stored in a grammar dictionary 111 to decide the connection between a candidate character string that is selected by a user or a character string that is directly inputted by the user and a character string that is stored in a character string buffer 112 in the preceding retrieval step. When a connectable state is decided by the part 110, a character string register part 113 stores the character string that is selected or directly inputted by the user in the buffer 112. If an unconnectable state is decided by the part 110, plural character strings if inputted in the buffer 112 are connected together into a compound word which is registered in a character string learning table 106. Then the buffer 112 is cleared.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character string prediction method for predicting a character string subsequent to an input character string based on a character string input by an input device, and a method of using the character string prediction method. Document creation device.

[0002]

2. Description of the Related Art Conventionally, in a document creation device such as a Japanese word processor, when inputting a character string using a handwritten character recognition device or the like, unlike a keyboard input, one character is input. Time was required, and the burden on the input person was large.

Therefore, in order to reduce the burden on the input person, when a part of a character string to be input is input, a character string following the character string is predicted based on the input character string. That method has been considered.

[0004] In response to this, a method called input prediction has been considered, for example, as reported in the literature (the 48th Annual Convention of the IPSJ 4J-11, etc.), the first few characters of which are reported. There is a method of searching a word dictionary at the time when is input, and extracting a character string (word) starting with a few characters.

[0005] In the input of a character string "kanji dictionary", for example, when the input person inputs "kanji", a "dictionary" is predicted as a character string following the "kanji" and finally " It outputs a “kanji dictionary”.

[0006]

The character string prediction as described above is performed by searching a word dictionary when the first few characters are input and obtaining a character string (word) starting with the few characters as a prediction candidate. Therefore, it is assumed that a desired word is registered in the word dictionary in advance. In this case, the word dictionary contains multiple commonly used words,
For example, a compound word such as “patent proposal” is not registered. Therefore, when entering a compound word that is not registered in the word dictionary, enter the first few characters of each word, such as “patent” or “proposal”, and enter multiple prediction candidates for each word. It was troublesome work to select the intended one from the

[0007] Further, it is naturally impossible to predict a character string not only for compound words but also for other unregistered words. For this reason, when such a word was input many times, it was very inconvenient.

The present invention has been made in view of the above points. Even if an unregistered word is input once, it can be predicted by the first few characters of the unregistered word from the next input. It is an object of the present invention to provide a character string predicting method capable of performing the method and a document creating apparatus using the character string predicting method.

[0009]

[Means for Solving the Problems]

(1) The present invention has a word dictionary storing words, and a character string learning table for storing character strings not registered in the word dictionary. Is searched as a predicted character string candidate from the word dictionary or the character string learning table, and the character string selected in the predicted character string candidate is stored in the character string buffer in the previous search. Performs connection determination with the stored character string, stores the selected character string in the character string buffer when it is determined that connection is possible, and stores multiple characters in the character string buffer when it is determined that connection is not possible. If the strings are stored, the character string obtained by combining the strings is registered as a compound word in the character string learning table, and then the character string buffer is cleared.

According to such a configuration, each connectable character string among the predictive character strings sequentially selected is learned as a compound word, and thereafter, the learned character string is searched for as one of the prediction targets. Therefore, even if a compound word is input, once it is input, prediction can be performed by the first few characters of the word from the next input.

(2) In the present invention, in the configuration of the above (1), connection determination between a character string selected from candidates of a predicted character string and a character string stored in the character string buffer in a previous search. Is performed based on the character type.

According to such a configuration, the connection is determined based on the character type. In this case, if the selected character string is composed of one character type other than Hiragana and has the same character type as the previous character string, connection is enabled, and other connections are disabled.

(3) According to the present invention, in the configuration of the above (1), there is provided a grammar dictionary storing grammar information for determining connection between character strings, and the grammar dictionary is selected from candidates of a predicted character string. A connection between a character string and a character string stored in the character string buffer in a previous search is determined based on the grammar information stored in the grammar dictionary.

According to such a configuration, the connection is determined based on the grammar information. In this case, the selected character string and the previous character string are compounded according to a grammatical rule prepared in advance, such as "noun and verb are not directly connected" and "no compound if particle is present". It is determined whether or not connection is possible as a word.

(4) The present invention has a word dictionary that stores words and a character string learning table that stores character strings that are not registered in the word dictionary. A character string having the input character string at the beginning is searched from the word dictionary or the character string learning table as a candidate for a predicted character string. If the search does not yield a candidate intended by the user, the user The input character string is registered in the character string learning table.

According to such a configuration, when the user directly inputs a character string, the character string is learned as an unregistered word, and thereafter, the learned character string is searched as one of the prediction targets. Therefore, even if an unregistered word is input, once it is input, prediction can be performed by the first few characters of the word from the next input.

(5) The present invention has a word dictionary that stores words and a character string learning table that stores character strings that are not registered in this word dictionary. A character string having the input character string at the beginning is searched as a predicted character string candidate from the word dictionary or the character string learning table, and a character string selected from the predicted character string candidates or directly input by a user. Determines the connection between the entered character string and the character string stored in the character string buffer in the previous search, and stores the selected character string or the directly input character string in the character string buffer when it is determined that connection is possible. If a plurality of character strings are stored in the character string buffer when it is determined that connection is impossible, a character string obtained by combining them is registered as a compound word in the character string learning table, and then the It is intended to clear the string buffer.

According to such a configuration, the connectivity between the character string selected from the prediction candidates or the character string directly input by the user and the previous character string is determined. Those character strings are learned as compound words, and thereafter, the learned character strings are searched for as one of the prediction targets. Therefore, even if a compound word including an unregistered word is input, once it is input, prediction can be performed by the first few characters of the word from the next input.

(6) The invention according to (5), wherein the character string selected from the candidates of the predicted character string or the character string directly input by the user and the character string buffer in the previous search are used. Is determined based on the character type.

According to such a configuration, the connection is determined based on the character type. In this case, if the selected character string or the directly input character string is composed of one character type other than hiragana and is of the same character type as the previous character string, connection is enabled, and other connections are disabled.

(7) According to the present invention, in the configuration of the above (5), there is provided a grammar dictionary in which grammar information for determining connection between character strings is stored, and the grammar dictionary is selected from candidates of a predicted character string. A connection determination between a character string or a character string directly input by a user and a character string stored in the character string buffer in a previous search is performed based on the grammar information stored in the grammar dictionary. I do.

According to such a configuration, the connection is determined based on the grammar information. In this case, for example, the selected character string or the directly input character string and the previous character string in accordance with a grammatical rule prepared in advance, such as "noun and verb are not directly connected" and "there is no compound word when particle exists". It is determined whether the character string can be connected as a compound word.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram illustrating a configuration of a document creation device according to an embodiment of the present invention. This device is a Japanese word processor having a character string prediction function. In FIG. 1, an input device 101 inputs a character string to be predicted. This input device 101
Is a device that can input a character string mixed with Chinese characters or kana by recognizing handwritten characters from coordinates obtained from a handwritten tablet or the like, and output input information as a character code. The output device 116 is, for example, a CRT (Cathod
It is a screen display device such as an e-ray tube) or an LCD (Liquid Crystal Display), and allows the user to interactively proceed with document creation work.

The input control unit 102 buffers character data input from the input device 101 in the input buffer 103 to create a character string, and sends the character string to the character string search unit 104 when the character string is to be predicted.

The word dictionary 105 stores headings, readings, and connection information for each word. The character string learning table 106 is composed of an arbitrary number of words, and includes the word dictionary 1
05 stores a character string that is not registered.
When the character string search unit 104 receives the input character string from the input control unit 102, the character string search unit 104 searches the word dictionary 105 or the character string learning table 106 for a character string having the input character string at the beginning, and predicts a character string obtained by the search. Character string candidates are stored in the candidate buffer 108.

The prediction candidate control unit 107 is a character string search unit 1
The predicted character string searched by the user 04 is displayed, the user performs a candidate selection process, and the predicted character string determined by the user is stored in the character string buffer 112. Candidate buffer 10
Reference numeral 8 denotes a memory for storing predicted character string candidates. The character string input unit 109 performs an input process in a case where the user directly inputs the remaining characters following the input character string without selecting a candidate of the predicted character string. The stored character string is stored in the character string buffer 112.

The grammar dictionary 111 stores grammar information for determining connection between character strings. The character string buffer 112 stores the predicted character string selected by the user from the candidates of the predicted character string, or the character string directly input by the user without obtaining the intended candidate. The connection determination unit 110 refers to the character type or the grammatical information stored in the grammar dictionary 111, and searches for a candidate character string selected by the user or a character string directly input by the user in the character string buffer 112 in the previous search. Performs connection determination with the character string stored in. In this case, for example, depending on the grammar dictionary, it can be set so that only those which become compound nouns can be connected.

The character string registration unit 113 includes a connection determination unit 110
If the connection is determined to be possible, the character string selected by the user or the character string directly input by the user is stored in the character string buffer 112, and if the connection is determined to be impossible, the character string buffer 112 is stored in the character string buffer 112. If there are a plurality of stored character strings, a character string obtained by combining them is registered as a compound word in the character string learning table 106, and then the character string buffer 112 is cleared.

The output control unit 114 includes an output buffer 115
Is displayed on the output device 116. The output buffer 115 is a memory for storing a character string to be displayed on the output device 116.

Next, the operation of the embodiment will be described. FIG.
Is a flowchart showing a flow of processing from character input to output of a predicted character string in the embodiment. Character data input from the input device 101 is input to the input control unit 10.
In step 2, the character string is formed by being buffered in the input buffer 103 until it becomes a prediction target (step 201).
To 202).

When the input character string is to be predicted (Yes in step 203), the character string stored in the input buffer 103 is sent to the character string search unit 104, and the character string search unit 10
In step 4, a character string having the character string at the beginning is searched from the word dictionary 105 or the character string learning table 106.
Predicted character string candidates obtained by this search are stored in the candidate buffer 108 (steps 204 and 205).

The timing at which the input character string is to be predicted (the timing at which the prediction process is performed) is, for example, when the prediction button is pressed by the user, or when a predetermined number of characters or time is reached.

After the search for the predicted character string is completed, the predicted candidate control unit 107 displays the predicted character string candidates and performs a user candidate selection process (step 206). At this time, if the user does not select any candidate and directly inputs a character string following the input character string, for example, because the candidate intended by the user was not obtained, the character string input unit At 109, the character string is input (steps 208 to 210). That is, the character string input unit 10
Reference numeral 9 denotes an input buffer 1 which stores a character string (the total number of characters constituting a desired word) directly input by the user through the input device 101.
03.

As described above, when a character string is directly input, the character string is determined as an unregistered word, and registered in the character string learning table 106 by the character string registration unit 113 (step 211). The character string selected by the user or the character string directly input is temporarily stored in the output buffer 115 as an output character string (step 21).
2).

Here, the prediction candidate control unit 107 checks the character string buffer 112, and if a character string has already been stored therein (Yes in step 213), the connection determination unit 1
10, the character string currently stored in the output buffer 115 (the character string selected or directly input from the candidates of the predicted character string) and the character string stored in the character string buffer 112 in the previous search. A connection determination process with a character string (the character string last stored in the character string buffer 112) is performed (step 214).

This connection determination process is for checking whether or not each character string can be connected as a compound word, and specifically, is stored in the grammar dictionary 111 or the character type of each character string as follows. This is performed based on the grammar information.

(A) In the connection determination method based on the character type, if the character string is composed of one character type other than Hiragana and has the same character type as the previous character string, connection is possible, and the other characters cannot be connected. And One character type other than hiragana is used because compound words are often input using only kanji or katakana, and hiragana characters are often input as particles, making it difficult to determine them as compound words.

(B) In the connection determination method based on grammar information,
For example, the character string can be connected to the previous character string as a compound word according to grammatical rules prepared in advance, such as "Noun and verb are not directly connected", "Particles are not compound words if they exist". It is determined whether or not.

When it is determined that connection is possible by such connection determination processing (Yes in step 215), the character string currently stored in the output buffer 115 (selected character string or The directly input character string) is stored at the end of the character string buffer 112 (step 216). On the other hand, when it is determined that connection is not possible (No in step 215), the prediction candidate control unit 107 checks the character string buffer 112, and if a plurality of character strings are stored therein (Yes in step 217). ), An instruction is issued to the character string registration unit 113, and a character string obtained by combining these is registered as a compound word in the character string learning table 106 (step 218), and then the character string buffer 112 is cleared (step 219).

In step 213, if no character string is stored in the character string buffer 112, the character string stored in the output buffer 115 (selected character string or direct input Is stored in the character string buffer 112 (step 2).
20).

Thereafter, the character string stored in the output buffer 115 is displayed on the screen by the output control unit 114 (step 221). Finally, the input buffer 103, candidate buffer 108, and output buffer 115 are initialized in preparation for the next input (step 222).

Next, the above-described character string prediction processing will be described with a specific example. First, a conventional character string prediction process will be described with reference to FIG. Conventionally, when a compound word “information processing society” not registered in the word dictionary 105 is input, first, as shown in FIG.
, And select “information” from the prediction candidates. Next, as shown in FIG.
“Process” is selected from the prediction candidates. Next, "gaku" is input as shown in FIG. 9C, and "society" is selected from the prediction candidates. As described above, when "Information Processing Society" is input, the above operation must be repeated for each word.

On the other hand, in the present invention, once the above operation is performed, a desired character string can be obtained only by inputting the first few characters of the compound word from the next input. FIG.
The character string prediction processing of the present invention will be described with reference to FIG. For example, when inputting a compound word “Information Processing Society” not registered in the word dictionary 105, first, as shown in FIG. Select "Information" from. At this time, “information” that is the selected character string is stored in the character string buffer 112.

Next, as shown in FIG. 7B, "processing" is input, and "processing" is selected from the prediction candidates. Here, the connection determination between the character string “information” already stored in the character string buffer 112 and the character string “process” selected this time is performed. As a result, it is determined that connection is possible, and the current selected character string “processing” is added to the end of the character string buffer 112.

Next, "gaku" is input as shown in FIG. 9C, and "society" is selected from the prediction candidates. Here, the connection determination between the character string “processing” stored at the end of the character string buffer 112 and the current selected character string “society” is performed, and the current selected character string “society” at the end of the character string buffer 112. Is added.

Next, as shown in FIG. 4D, "to" is manually input, and an auxiliary word "to" is selected from the candidate character strings.
Is selected. Since this selected character string "to" cannot form a compound word (compound noun), it is determined that connection is impossible. At this time, since the character string buffer 112 stores a plurality of character strings “information”, “processing”, and “society”, the “information processing society” combining these character strings is stored in the character string learning table 106. be registered. In addition, the character string buffer 112 is cleared by the input of the character string “was” that cannot be connected.

As described above, with the input of a character string that cannot be connected as a trigger, a plurality of character strings stored in the character string buffer 112 before that are registered in the character string learning table 106 as one compound word. . As a result, when “information” is input again as shown in FIG. 11E, “information processing society” is obtained as a prediction candidate using the character string learning table 106.

In the present invention, a character string directly input by the user once is also listed as a prediction candidate from the next input. A specific example at this time is shown in FIG. For example, FIG.
As shown in FIG. 5, when the user inputs “cha” with the intention of a character string such as “Charles”, but there is no “Charles” in the prediction candidate, that is, when the word is not registered in the word dictionary 105, As shown in (b), the user directly inputs "Charles".

Here, the character string directly input by the user is stored in the character string buffer 112, and the character string is registered in the character string learning table 106 as an unregistered word. As a result, when "ch" is input again as shown in FIG. 9C, "charles" is obtained as a prediction candidate using the character string learning table 106.

Further, in the present invention, even when a compound word including an unregistered word such as "Prince Charles" is input, it can be predicted. This case can be realized by storing “Charles” and “Prince” in the character string buffer 112 and then performing a connection determination between the two.

The present invention can be applied not only to the input of handwritten characters as described in the above embodiment, but also to the input by a keyboard. Even in such a case, the input operation of the user is performed in the same manner as described above. Time and labor can be reduced, and document creation efficiency can be improved.

[0052]

As described above, according to the present invention, each connectable character string among the predictive character strings sequentially selected is learned as a compound word, and thereafter, the learned character string is set as one of the prediction targets. Since the search is performed, even if a compound word is input, once it is input, prediction can be performed by the first few characters of the word from the next input. As a result, user input operations can be reduced, and the efficiency of document creation can be improved.

According to the present invention, when a user directly inputs a character string, the character string is learned as an unregistered word, and thereafter, the learned character string is searched as one of the prediction targets. Therefore, even if an unregistered word is input, once it is input, it can be predicted from the next input using the first few characters of the word. This allows
User input operations can be reduced, and the efficiency of document creation can be improved.

According to the present invention, the connectivity between the character string selected from the prediction candidates or the character string directly input by the user and the previous character string is determined. Since those character strings are learned as compound words, and the learned character strings are searched as one of the prediction targets thereafter, even if a compound word including an unregistered word is input, once it is input, the next time it is input, Input makes it possible to make predictions using the first few characters of the word. This allows
User input operations can be reduced, and the efficiency of document creation can be improved.

[Brief description of the drawings]

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

FIG. 2 is an exemplary flowchart showing the operation of a character string prediction process in the embodiment.

FIG. 3 is a specific example for explaining a conventional character string prediction process.

FIG. 4 is a specific example for explaining a character string prediction process according to the present invention.

FIG. 5 is a specific example for explaining the character string prediction processing of the present invention.

[Explanation of symbols]

 Reference Signs List 101 input device 102 input control unit 103 input buffer 104 character string search unit 105 word dictionary 106 character string learning table 107 prediction candidate control unit 108 candidate buffer 109 character string input unit 110 connection determination unit 111 grammar dictionary 112 character string buffer 113 character string registration unit 114 output control unit 115 output buffer 116 output device

Claims (14)

[Claims] A word dictionary storing words and a character string learning table for storing character strings that are not registered in the word dictionary are provided. The character string in the part is searched as a predicted character string candidate from the word dictionary or the character string learning table, and the character string selected in the predicted character string candidate and stored in the character string buffer in the previous search. The connection determination is performed with the connected character string. When the connection is determined to be possible, the selected character string is stored in the character string buffer. When the connection is determined to be impossible, a plurality of character strings are stored in the character string buffer. If stored,
A character string predicting method comprising: registering a character string obtained by combining them as a compound word in the character string learning table; and then clearing the character string buffer. 2. A connection determination between a character string selected from candidates of a predicted character string and a character string stored in the character string buffer in a previous search is performed based on a character type. 1. The character string prediction method according to 1. 3. A grammar dictionary for storing grammar information for determining connection between character strings, and a character string selected from candidates of a predicted character string and stored in the character string buffer in a previous search. 2. The character string prediction method according to claim 1, wherein the connection determination with the selected character string is performed based on the grammar information stored in the grammar dictionary. 4. It has a word dictionary storing words, and a character string learning table for storing character strings not registered in the word dictionary. When a character string to be predicted is input, the input character string A character string in a part is searched from the word dictionary or the character string learning table as a candidate for a predicted character string. If the search does not yield a candidate intended by the user, a character string directly input by the user Is registered in the character string learning table. 5. A word dictionary that stores words and a character string learning table that stores character strings that are not registered in the word dictionary. When a character string to be predicted is input, the input character string is read first. A character string in a part is searched as a predicted character string candidate from the word dictionary or the character string learning table, and a character string selected from the predicted character string candidates or a character string directly input by a user is searched. Performs connection determination with the character string stored in the character string buffer in the previous search, and if it is determined that connection is possible, stores the selected character string or the directly input character string in the character string buffer and disables connection If it is determined that a plurality of character strings are stored in the character string buffer, a character string obtained by combining them is registered as a compound word in the character string learning table, and then the character string buffer is stored. String prediction method which is characterized in that clear. 6. A connection determination between a character string selected from candidates of a predicted character string or a character string directly input by a user and a character string stored in the character string buffer in a previous search is set as a character type. 6. The method according to claim 5, wherein
The described character string prediction method. 7. A grammar dictionary storing grammar information for determining connection between character strings, wherein a character string selected from candidates of a predicted character string or a character string directly input by a user. 6. The character string prediction method according to claim 5, wherein connection determination with a character string stored in the character string buffer in a previous search is performed based on the grammar information stored in the grammar dictionary. 8. An input means for inputting a character string to be predicted, a word dictionary storing words, a character string learning table storing character strings unregistered in the word dictionary, Character string search means for searching a character string having the first character string as a predicted character string candidate from the word dictionary or the character string learning table, a character string buffer for storing the determined character string, Connection determination means for determining a connection between a character string selected from the candidates of the predicted character string searched by the character string search means and a character string stored in the character string buffer in a previous search; When the determination unit determines that connection is possible, the selected character string is stored in the character string buffer, and when it is determined that connection is not possible, a plurality of character strings are stored in the character string buffer. And a character string registering means for registering a character string obtained by combining them as a compound word in the character string learning table and then clearing the character string buffer. 9. The connection determining unit performs a connection determination between a character string selected from candidates of a predicted character string and a character string stored in the character string buffer in a previous search based on a character type. 9. The document creation device according to claim 8, wherein: 10. A grammar dictionary storing grammar information for determining connection between character strings, wherein the connection determination unit determines a character string selected from candidates of a predicted character string by a previous search. 9. The document creation apparatus according to claim 8, wherein connection determination with a character string stored in the character string buffer is performed based on the grammar information stored in the grammar dictionary. 11. An input means for inputting a character string to be predicted, a word dictionary storing words, a character string learning table storing character strings not registered in the word dictionary, Character string search means for searching a character string having the character string at the beginning as a candidate for a predicted character string from the word dictionary or the character string learning table; and a candidate intended by the user is obtained by the character string search means. A character string registering means for registering a character string directly inputted by a user in the character string learning table when there is no character string. 12. An input means for inputting a character string to be predicted; a word dictionary storing words; a character string learning table storing character strings not registered in the word dictionary; Character string search means for searching a character string having the first character string as a predicted character string candidate from the word dictionary or the character string learning table, a character string buffer for storing the determined character string, Connection between a character string selected from among the predicted character string candidates searched by the character string search means or a character string directly input by the user and a character string stored in the character string buffer in a previous search. Connection determination means for performing determination; storing the selected character string in the character string buffer when the connection determination means determines that connection is possible; and determining that connection is not possible. If a plurality of character strings are stored in the character string buffer, a character string registration unit that clears the character string buffer after registering a character string obtained by combining them as a compound word in the character string learning table. A document creation device, comprising: 13. The connection judging means according to claim 1, wherein the character string selected from the candidates of the predicted character string or the character string directly input by the user and the character string stored in the character string buffer in the previous search. 13. The document creation apparatus according to claim 12, wherein the connection determination is performed based on a character type. 14. A grammar dictionary storing grammar information for determining connection between character strings, wherein the connection determining means is a character string selected from candidates of a predicted character string or directly by a user. 13. The document according to claim 12, wherein connection determination between an input character string and a character string stored in the character string buffer in a previous search is performed based on the grammar information stored in the grammar dictionary. Creating device.