JPS63182735A - Voice input word processor - Google Patents

Abstract

PURPOSE:To extremely reduce the labor and the time needed for the key input of an operator and to produce a document at a high speed, by performing the display and the final selection of the voice input for each character string. CONSTITUTION:A voice/KANA (square form of Japanese syllabary) code converting part 11 recognizes the voice input for each monosyllable and converts each monosyllable into a KANA code consisting of plural candidates. A character code string selecting part 12 combines continuously those candidates included in each converted KANA code in the order of conversion for production of a character code string to collate this code string with another character code string under registration and disuses the discordant codes to select a character code string consisting of a single or plural candidates. A display part 16 displays one of those selected character code strings as a single character string. A document producing part 11 uses the designated one displayed character strings to produce documents. Thus the voice input data are displayed and finally selected for each character string. As a result, the labor and the time needed for key operations of an operator omitted and therefore the documents are produced at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a voice input word processor used in office processing systems and the like.

2. Description of the Related Art Recently, word processors that create documents using voice input have been developed.

In such voice-input word processors, the operator utters single syllables or letters one by one while the machine automatically recognizes them and displays the first candidate for the recognition result.If the first candidate for the recognition result is unintended, the recognition result can be changed by pressing the correction key. The second. A method is used in which the third candidates are displayed one after another.

Problems to be Solved by the Invention In the conventional voice-powered devices described above, it is generally difficult to increase the speech recognition rate because the input speech varies due to various fluctuation factors such as individual differences between operators and surrounding noise. Therefore, there is a problem that the number of operations of the correction key by the operator increases and the processing speed decreases.

Composition of the Invention Means for Solving the Problems The speech input word processor according to the present invention recognizes speech input monosyllable by monosyllable and converts each monosyllable into a kana code consisting of a plurality of candidates. The conversion section continuously combines multiple candidates included in each of the converted kana codes to create a character code string, which is compared with the registered character code string, and those that do not match are discarded. a character code string selection section for selecting a character code string consisting of one or more candidates; a display section for displaying one of the selected character code strings as a character string; By displaying voice input data and making final selections in character string units, the operator can save key operations and speed up document creation. is configured to do so.

Hereinafter, the operation of the present invention will be explained in detail together with examples.

Embodiment FIG. 1 is a block diagram showing the configuration of a voice input word processor according to an embodiment of the present invention.

In the figure, 10 is a control unit, 11 is a voice/kana code conversion unit,
12 is a kana code string selection section, 13 is a kana code string dictionary,
14 is a character string buffer, 15 is a display control unit, 16 is a CR
T (cathode ray tube), 17 a kana/kanji converter, 18 a kanji dictionary, 19 a document file, and 20 a keyboard.

The voice/kana code converter 11 recognizes the voice uttered by the operator on a monosyllable basis and converts it into the corresponding kana code. At this time, for one single syllable, from the first candidate to the nth
The n kana codes up to the candidates are selected and each is transferred to the kana code string selection section 12 via the control section IO.

That is, as illustrated in FIG. 2, when the operator utters the first monosyllables rA, J, the recognition result is a group of n kana codes r from the first candidate to the nth candidate.
azJ, rat□''...r a , fiJ are generated by the speech/kana code conversion section 11, and each is transferred to the kana code string selection section 12 via the control section 10.

Next, when the monosyllable "A2" is uttered, n kana codes "agl" + "az□" ... "al" from the first candidate to the nth candidate are generated as a recognition result. is,
It is transferred to the kana code string selection section 12. Similarly, each time the operator utters a monosyllable, n kana codes from the first candidate to the nth candidate are generated by the speech/kana code converter 11 and transferred to the kana code string selector 12.

In the example in Figure 2, the voiceless state (operator's pause state) continues for a predetermined period of time after the single syllable "A1" is uttered.
shall continue. When the control unit IO detects the silent state since no subsequent kana code is transferred from the voice/kana code conversion unit 11 for a predetermined period of time, the control unit IO causes the kana code string selection unit 12 to start the kana code string selection process.

When starting the selection process, the kana code string selection unit 12 selects a combination of two kana codes by combining multiple candidates included in each of the first monosyllable "A1" and the second monosyllable "A2". nxm kana code strings al1221 + azg” ” ” alla
2nalZa21+a12a! 2°""a12a2n is created.

Next, the kana code string selection unit 12 compares each of the nXm kana code strings created as described above with the kana code strings registered in the kana code string dictionary 13, and selects the ones that do not match. will be discarded.

The kana code string selection unit 12 newly combines n candidates included in the third monosyllable "A3" with the remaining kana code strings without being discarded, thereby creating a combination of three kana codes each. Create multiple kana code strings consisting of .

Subsequently, each newly created kana code string is compared with the kana code strings registered in the kana code string dictionary 13, and those that do not match are discarded. Then, the n candidates included in the fourth monosyllable are newly combined with the kana code string consisting of the three kana codes that remained without being discarded, resulting in a character code string consisting of four kana codes. are created, each of these is compared with the kana code strings registered in the kana code string dictionary 13, and those that do not match are discarded.

In this way, the kana code string selection section 12 selects the voice/
A plurality of candidates included in each of the kana codes converted by the kana code converter 11 are successively combined in the converter to create a character code string, which is compared with the kana code string currently registered in the kana code string dictionary 13. , a character code string consisting of one or more candidates is selected by discarding those that do not match.

The kana code strings selected by the kana code string selection unit 12 are ranked as first candidate, second candidate, third candidate, etc. by an appropriate method, and are stored in the character string buffer 14 according to this ranking. At the same time, the first candidate kana character string is automatically displayed in a predetermined area on the CRT 16 while being converted into a character string by the display control unit 15.

If the first candidate kana character string displayed on the CRT 16 is contrary to expectations, the operator presses the keyboard 2.
Press the correction key above 0. Accordingly, a kana character string corresponding to the second candidate kana code string currently stored in the character string buffer 14 is newly displayed on the CRT 16 in place of the first candidate.

Before pressing the correction key, or by pressing the correction key several times, the operator confirms that the expected kana character string is CR
When displayed on T16, press the kana/kanji conversion key on the keyboard 20. When the control unit IO detects the depression of this key, it starts the kana/kanji conversion unit 17, refers to the kanji dictionary 18, converts the kana code string to the first candidate kanji-mixed character string, and displays the display unit 15. via CRT16
to be displayed. If the operator finds that the character string containing kanji does not meet his expectations, he presses the conversion key again to display the next candidate character string containing kanji. If the character string containing Kanji characters converted in this way is what the operator expects, the operator presses the no-conversion key to confirm the input, and then transfers it to the document file 19. The sentences created in this manner based on human voice are stored in the document file 19 as sentences containing kanji and kana.

3 and 4 are flowcharts for supplementary explanation of the processing procedure by the control section 10. FIG.

The flowchart in FIG. 3 shows the voice/kana code converter 1
1 and transfers the kana code to the kana code string selection section 12 (steps 31 to 33), detects a silent state for a predetermined period and causes the kana code string selection section 12 to start selecting a kana code string (steps 34 to 33). 37), receives the selected kana code string, transfers it to the character string buffer 14, and transfers it to the CRT1.
6 (steps 38 to 40).

In the flowchart of FIG. 4, with the kana code string stored in the character string buffer 14, the correction key (step 43), the conversion key (step 45),
Processing when the no-conversion key (step 47) is pressed is shown in steps 44, 46, and 48, respectively.

The processing of FIGS. 3 and 4 is typically processed as a separate task under O8.

Next, the explanation of the kana character string selection process by the kana character string selection unit 12 will be supplemented using a specific example.

When the operator tries to add the kanji ``University'' to a document he is creating, he ends up with four monosyllables ``DA''.

After saying “I”, “gaJ, r<J” almost continuously,
It is assumed that the user waits for the kana code string to be displayed while maintaining a silent state.

In the voice/kana code converter 11, each of the first and second digits is converted as follows. Second. It is assumed that conversion to a kana code consisting of the third candidate has been performed.

When the kana code string selection unit 12 is activated by a predetermined silent state, it first selects the first kana code, the second candidate, and the third candidate. By combining the first to third candidates for the second kana code, nine kana code strings are created: damikami temidani kani deni tai ga dei.

Next, the kana code string selection unit 12 compares each of the nine kana code strings with the kana code strings currently registered in the kana code string dictionary I3. - For example, for a character string that includes "kai",
It is assumed that the five types of kana code strings are registered in the dictionary 13: Gaikai (general, town), Gaikai (outside sea, outer world), Kaikaku (outer corner, outer area), Kaikatsu (general), and Gaikaan (exterior, external appearance, general appearance). . Therefore, one of the nine kana codes created, "kai", matches and remains as a candidate. In addition,"
Kanji characters in parentheses, such as "", are added for reference, and do not necessarily mean that a corresponding kanji code is provided in addition to the kana code.

The other kana code strings are also compared with the kana code string dictionary 13 in the same way, and the following selection results (
0: in the dictionary, ×: not in the dictionary) is obtained.

Dami Ogami × Demi × Dani 0 Kani O deni × Dai 0 Gai Odei O In this way, the remaining six kana code strings include the first of the third recognized kana character. By combining the candidates to the third candidate, the following 18 kana code strings are obtained.

Damiga × Daniga × Daiga Odamiga ×
Daniba × Daiba 0damido × Danido ×
Daido O crab ga × ga ga × deiga
× Crab × Gaiba × Deiba × Crab ×
Gaido 0 Deido ×In this way, only four kana code strings remain.

If we further combine the first to third candidates of the fourth converted kana code to these remaining kana code strings, we get
The following 14 character code strings are obtained.

Daigaku O Daibaku × Daidoku 0 Daibaku ×
Daibafu × Gaidofu × Daigami × Daibami × Daidomi × Gaidoku Ogaidofu × Gaidomi × In this way, three kana code strings are ultimately left as candidates.

Here, regarding the final three remaining kana code strings,
Calculating the sum of the rankings of the candidates given when converting each constituent kana code, the candidate ranking is approximately 131 l...6 (total) kana code
Daidoku candidate ranking 1331...8 (sum total) kana code
The highest candidate ranking is 2331...9 (total sum) Therefore, if the first, second, and third candidates are selected in descending order of the total sum of candidate rankings, the results are as follows.

First candidate, second candidate, third candidate, first candidate, etc. Above, we have exemplified a configuration in which a silent state for a predetermined period of time is automatically detected and the kana code string selection process is started. However, the kana code string selection process for the converted kana code may be performed in parallel with the conversion process from monosyllables to kana codes, and the first candidate may be displayed as soon as a silent state for a predetermined period of time is detected. A configuration may also be adopted in which selection and display processing are started in response to manual designation of appropriate keys.

Further, by slightly expanding the functions of the voice/kana code conversion document 11, it is possible to make simple and fixed commands such as "correction" and "conversion" by key operations using the operator's voice.

Effects of the Invention As explained in detail above, the voice input word processor of the present invention is configured to display and final select voice input in character string units, so that the operator's manual effort is greatly reduced. This has the effect of speeding up document creation.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a speech input word processor according to an embodiment of the present invention, and FIG. 2 shows kana characters recognized by the speech/kana code converter 11 of FIG. 1 and transferred to the kana code string selector 12. A conceptual diagram illustrating the structure of the code, and FIGS. 3 and 4 are flowcharts for supplementary explanation of the operation of the control unit 10 shown in FIG. 1. lO...control unit, 11...voice/kana code conversion unit, 12...kana code string selection unit, 13...kana code string dictionary, 14...buffer, 15...display control unit , 17... Kana/Kanji conversion section, 18... Kanji dictionary,
19...Document file.

Claims (3)

[Claims] (1) A speech/kana code converter that recognizes speech input monosyllable by monosyllable and converts each monosyllable into a kana code consisting of multiple candidates; and a speech/kana code conversion unit that recognizes the speech input monosyllable by monosyllable and converts each monosyllable into a kana code consisting of multiple candidates, and the multiple candidates included in each of the converted kana codes are consecutively connected in the conversion order. a character code string selection unit that selects a character code string consisting of one or more candidates by collating the character code strings that are currently being registered and discarding those that do not match; The present invention is characterized by comprising a display unit that displays one of the selected character code strings as a character string, and a document creation unit that creates a document using a specified one of the displayed character strings. Voice input word processor. (2) The voice input word processor according to claim 1, wherein the display of the selected character string is automatically started when voice input is interrupted for a predetermined period of time. (3) The display of the selected character strings has a candidate ranking according to the sum of the candidate rankings of the kana codes that constitute each character string, and can be displayed in an order according to this candidate ranking. A voice input word processor as claimed in claims 1 and 2.