JPH0797370B2 - Character / speech input conversion method - Google Patents

Description

TECHNICAL FIELD OF THE INVENTION The present invention is a character / speech input conversion suitable for efficiently creating, for example, a Japanese sentence mixed with kana / kanji from a character string input from a voice recognition device or a character reading device. method.

[Technical background of the invention and its problems]

In recent years, as the efficiency of office work has advanced, a large number of document creation devices centering on word processors have been introduced into offices.

Most word processors of this type are of a type in which reading information of a created sentence is input as a kana character string by operating a keyboard, for example. However, the kana key operation is more complicated than the English type key operation. Therefore, the Roman alphabet input method and other input methods have been proposed, but the problem of the operability has not been essentially improved.

Therefore, as input means of the created sentence information, application of handwritten character input using a character reading device or voice input using a voice recognition device has been considered.

However, it is almost impossible to achieve a recognition rate of 100% in this kind of recognition device. Therefore, it is undeniable that, for example, when inputting a sentence on one line, erroneous input corrections need to be performed a plurality of times. Specifically, it cannot be denied that the input voice is erroneously recognized due to personal fluctuation factors of utterance,
In addition, it cannot be denied that erroneous recognition may occur due to pattern deformation of handwritten characters. Therefore, it is necessary to correct the input error each time the input character is erroneously recognized.

Therefore, even if the recognition device is used together, a lot of labor is required to correct the input error, and there is a problem that the document creation efficiency cannot be improved in the end.

[Object of the Invention]

The present invention has been made in consideration of such circumstances, and an object thereof is to surely obtain a desired conversion result (input information) from a character string input from a voice recognition device or a character reading device, For example, it is to provide a character / speech input conversion method suitable for efficiently creating a Japanese sentence mixed with kana and kanji.

[Outline of Invention]

The present invention finds, for example, a plurality of kana-kanji conversion candidates for a sequence of input character candidates obtained by a recognition device, and when one of them is selected, the conversion candidate and the input character candidate corresponding thereto are selected. The column is registered in, for example, the provisional dictionary unit or the personal dictionary unit of the conversion dictionary, and thereafter, the registered information is effectively used to efficiently perform the conversion process on the input character candidate column.

〔The invention's effect〕

Thus, according to the present invention, even if an input voice or an input character that is easily misrecognized is registered in the conversion dictionary in association with a string of input character candidates and a conversion result, once the conversion candidate is selected. By doing so, even if similar erroneous recognition occurs for the same input voice or input character, a correct (desired) conversion result can be obtained regardless of the erroneous recognition.

As a result, it is possible to greatly reduce the selection work for a plurality of conversion candidates, simplify it, and improve the efficiency of the text creation work.

Example of Invention

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a Japanese sentence creation device by voice input configured by applying the embodiment method.

The voice recognition unit 1 recognizes the voice input by voice, for example, in syllable units, and outputs the recognition candidates. When a plurality of recognition candidates are obtained for one syllable, the voice recognition unit 1 ranks and outputs the plurality of recognition candidates, for example, according to the similarity to the recognition dictionary that has obtained the recognition candidates.

The buffers 2 and 3 create a sequence of input character candidates from the recognition candidates of the input voice that have been subjected to the recognition processing by the voice recognition unit 1. That is, when the voice recognition unit 1 sequentially obtains recognition candidates "so, u, shi (chi)" for the input voice "so-u-chi", the input buffer 2 is set to the input buffer 2 shown in FIG. ), The input speech recognition candidates are sequentially stored. The buffer 3 obtains “Soushi” as the sequence of the input character candidates of the first candidate as the combination of the recognition candidates stored in the buffer 2 as shown in FIG. We are looking for "souchi" as a candidate line.

The control unit 4 controls the extraction processing of the input character candidate string by the buffers 2 and 3 described above.

Then, the kana-kanji conversion unit 5 searches the conversion dictionary 6 for a conversion candidate for the sequence of input character candidates obtained in the buffer 3, and stores the conversion result (conversion candidate) in the page buffer memory 7. .

The conversion dictionary 6 is composed of, for example, a resident dictionary unit 6a, a personal dictionary unit 6b, and a temporary dictionary unit 6c. The resident dictionary unit 6a is a standard prepared in advance as a conversion result of Kanji (Kanji string) for a character string indicating reading, and the personal dictionary unit 6b is used by the user of the device, for example, personally. This is an area for registering frequently used kanji (kanji string) and a character string indicating its reading in association with each other.

The provisional dictionary section 6c is a dictionary area that is provisionally used each time an operation is performed, and registers information of words (conversion candidates) selected and designated from among homophones that are peculiar to Japanese, for example. Is. In other words, when the same conversion candidate is selected, the word (conversion candidate) previously selected is often selected again, so this is the dictionary area used for preferentially outputting this. .

The kana-kanji conversion unit 5 searches the conversion dictionary 6 having the above-described configuration, for example, according to the procedure shown in FIG. 3, and obtains conversion candidates for a string of input character candidates.

Then, the conversion candidates obtained by the kana-kanji conversion unit 5 are displayed on the display unit 8 under the control of the control unit 4. When a plurality of conversion candidates are obtained for the same input character string, the control unit 4 displays, for example, the conversion candidate obtained in the first place, and another conversion candidate other than the conversion candidate is displayed. Display that there are also candidates. The display of the next candidate is performed by, for example, changing the display brightness of the displayed conversion candidate, displaying the conversion candidate in reverse video, or underlining.

When it is necessary to select a plurality of conversion candidates, that is, when it is desired to obtain a conversion candidate different from the conversion candidate displayed in the first place, the keyboard 9 is instructed to display another conversion candidate, and the desired conversion candidate is displayed. This is performed by selecting and instructing the conversion candidate to be converted.

When a conversion candidate is selected, the selected conversion candidate and the corresponding input character string are registered in the conversion dictionary 6. This dictionary registration is performed by the temporary dictionary unit 6c.
Is performed for the personal dictionary unit 6b as necessary.

Now, the search of the conversion dictionary 6 for the string of input character candidates is performed as follows, as shown in the procedure in FIG.

First, it is determined whether or not the input character string includes a plurality of candidates (step a). When only a single input character candidate string is obtained, the search parameters M and N of the conversion dictionary 6 are initially set to (N = 1) and (M = N) (step b,
c). When there are a plurality (n ways) of candidates for the input character string, initial setting is made to (N = n) and (M = N) (steps d and e). Incidentally, M is a search control parameter for the resident dictionary 6a and personal dictionary 6b, and N is a search control parameter for the temporary dictionary 6c.

Therefore, first, the resident dictionary 6a and the personal dictionary 6b are sequentially searched according to the parameter M, and it is sequentially searched whether or not the conversion word having the column of the input character candidates as a headline is registered. Then, each time a corresponding registered word is found, the registered word is sequentially stacked on the stack as a conversion candidate (step f). Then, the parameter M is decremented (M ← M-1), its value is determined, and the dictionary search process is repeated until the parameter M becomes (0) (step g). By repeating this, a dictionary search is performed for each of a plurality of input character candidate strings.

In this example, first, a dictionary search is performed for a sequence of input character candidates of "Soshi" to obtain conversion candidates of, for example, "foundation" and "sagami", and these are put on the stack. After that,
A dictionary search is performed on the sequence of input character candidates such as "Souchi" to obtain a conversion candidate such as "apparatus", which is stacked on the stack. As a result, three conversion candidates, "foundation,""paper," and "device," are required for the stack.

In this case, the conversion candidates are stacked on the stack in the order of “foundation”, “grass paper”, and “device”.

Then, the temporary dictionary 6c is searched according to the parameter N (steps h and i). Then, when a registered word corresponding to the input candidate string is present in the temporary dictionary 6c, the registered word is set at the top of the stack and the stack is reloaded. As a result, the conversion candidate registered in the temporary dictionary 6c is obtained as the highest priority candidate for the input character string.

Specifically, when "apparatus" is registered in the provisional dictionary 6c as "souchi" reading, "apparatus" is transferred to the top of the stack, and "apparatus""found" is added to the stack.
The conversion candidates will be piled up in the order of "Soshi". As a result, the conversion candidate "device" is preferentially output to the input character string.

Thus, in this device, once the input character string is selected and designated, the conversion candidates and the input character string selected thereby are registered in the provisional dictionary 6c. When a character string is given, the previously selected conversion candidate is preferentially output. Therefore, even when the input character string is indicated by a plurality of candidates such as “Soushi” and “Souchi”, especially when the input character string is erroneously recognized as “Soushi” as the first candidate. Also, it becomes possible to efficiently obtain the correct (desired) conversion candidate of the erroneous input character string from the next candidate "Souchi".

In this example, as shown in FIG. 4 (a), the conversion result “device” is stored in association with the character string “souchi” indicating the correct reading, and the reading is used as a candidate for the input character string. When it is obtained, the correct conversion candidate is preferentially obtained.

However, as shown in FIG. 4B, the input character string corresponding to the selected conversion candidate may be given as the character string to be developed as the first candidate of the recognition result.

Specifically, a conversion candidate "device" may be registered in association with the input character candidate string "soushi". By doing so, it is possible to preferentially obtain a correct (desired) conversion candidate by simply performing a dictionary search on the input character candidate sequence expanded to the first place for the input character string. Become.

In this way, when the recognition device is used as the information input means, erroneous recognition may occur due to the personality of the input person or the recognition processing characteristic of the recognition device. However, since the original conversion candidate can be easily and automatically obtained, the input work can be greatly simplified. Therefore, there are practically great effects such that the efficiency of the document creation work can be effectively achieved.

It should be noted that the present invention is not limited to the above-described embodiments, but can be widely applied to devices other than the document creation device. In addition, although the description has been given here using voice input as an example, the present invention can be similarly applied to document creation by handwritten character input using OCR. In short, the input conversion method according to the present invention can be applied to various information input processing, and various modifications can be carried out without departing from the scope of the invention.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a document creating apparatus to which an embodiment system of the present invention is applied, FIG. 2 is a diagram showing a configuration of a buffer and its operation in the embodiment apparatus, and FIG. 3 is a conversion dictionary search processing procedure. FIG. 4 is a diagram showing a form of registration of the selected conversion candidate in the provisional dictionary. 1 ... Voice recognition section, 2,3 ... buffer, 4 ... control section,
5 ... Kana-Kanji conversion unit, 6 ... conversion dictionary, 6a ... resident dictionary, 6b ... personal dictionary, 6c ... provisional dictionary, 7 ... page buffer memory, 8 ... display unit, 9 ... keyboard.

Claims (2)

[Claims] 1. A recognition means for recognizing an input pattern composed of an input character or voice and obtaining a character candidate for the input pattern, and a character candidate obtained by this recognition means are combined to generate a string of input character candidates. Candidate string generation means, conversion means for collating the input character candidate columns generated by the candidate string generation means with a conversion dictionary to obtain a conversion result for the input character candidate column, and the conversion result on the display device. Display control means for displaying, selection means for selecting one of the conversion candidates when the displayed conversion result includes a plurality of conversion candidates, the selected conversion candidate, and the conversion candidate A combination means for obtaining a combination of input character candidate strings, and a set of the conversion candidates and the input character candidate strings obtained by the combination means are registered in the conversion dictionary. And a unit that controls the combination candidate to combine one of the plurality of input character candidate strings with the conversion candidate when the plurality of input character candidate strings are acquired by the candidate string generation unit. Character / speech input conversion method characterized by 2. A string of input character candidates registered in the conversion dictionary together with the selected conversion candidate is a character string indicating correct reading for the conversion candidate, or an input character expanded to the first position in the candidate string generation means. The character / speech input conversion method according to claim 1, which is composed of a sequence of candidates.