JPH07120253B2 - Text input device by voice - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to a text input / processing / converting device (hereinafter, word processor) by voice input, and particularly improves efficiency of confirmation / correction and reduces time required therefor. The present invention relates to a text input / processing / conversion device that is suitable.

[Background of the Invention]

The confirmation and correction methods in the conventional word processor are mainly the visual confirmation of the input result displayed on the screen (hereinafter, display) and the correction after the position is designated by the manual cursor. FIG. 1 shows a processing flow of a conventional confirmation / correction method. First, visually confirm the input result displayed on the display, and if there is an input or recognition error, manually move the cursor to the error position and then calibrate correction, deletion, insertion, etc. I do. The visual confirmation is troublesome because it is necessary to alternately see the original and the display, and there is a drawback in that the confirmation requires a long time and is inefficient. Further, there is a drawback in that the movement of the cursor by hitting the cursor keys is time-consuming and inefficient.

[Object of the Invention]

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problems, to provide a word processor which has a high efficiency of confirmation and correction and which requires a reduced time.

[Outline of Invention]

In order to achieve the above object, in the present invention, among the recognition results of the input voice, an uncertain part is detected and the uncertain part is displayed differently from other recognition results. And the correction cursor was positioned at the uncertain place.

Specifically, the voice-based text input device according to the present invention is
Input means for inputting voice, voice recognition means for analyzing / recognizing the input voice, display means for displaying the result recognized by the voice recognition means, and when the recognition result of the voice is uncertain In the recognition result displayed on the display means, the uncertain portion is displayed in a distinguishable manner, and a control unit for controlling to position the correction cursor at the uncertain portion is constituted. It

Detection of an uncertain part of the voice recognition result is performed by providing the voice recognition means with a predetermined matching distance reference, and when the voice recognition result shows the reference value or more, or when the voice recognition means obtains the voice matching. Among the distances, when the ratio between the first-ranking matching distance and the second-ranking matching distance is less than a predetermined reference, the voice recognition result is detected as uncertain.

As a pattern recognition system relating to a display method, there is, for example, JP-A-57-211200.

Example of Invention

Hereinafter, the principle of the present invention will be described in detail. FIG. 2 shows a processing flow of a text input method incorporating a display, confirmation and correction method according to the present invention. For example, the voice input in phrase units is voice-recognized, and the recognition result is displayed on the screen. At this time, a display method is used so that an uncertain portion (voice) of the recognition result or a portion that is likely to be erroneously recognized can be distinguished from other portions. Figure 3 shows the display method 1
This is an example. In case of erroneously recognizing two places in input voice / onsen // (/ i / → / hi /, / yo / →
/ O /) was assumed. A part that is erroneously recognized is detected as a recognition uncertainty part from the matching distance value of the recognition result, and that part is changed in color or marked to give a special display. The next displayed result is mainly recognized mainly in the part where the mark is attached, and if there is no error in the input result, the process returns to the next Kanner-Kan conversion. If there is an error in the input result, the correction cursor is automatically positioned in advance to the position where it seems to have been incorrect (Fig. 3).
If the point is really an error, correct the result. If not, move to the next position and continue the same process. As a result, the location of the error can be easily found, and the position of the error correction cursor moves quickly, so that efficient confirmation and correction can be performed. As a method of correction, first, the next candidate display of the recognition result, then re-input by voice, and finally by key input can be considered. When there is no error in the input result, kana-kanji conversion processing, confirmation of the result, and correction processing are performed. The above processing is continued until the input of all sentences is completed. A form in which the display / correction method according to the present invention is incorporated after Kana-Kanji conversion is also conceivable.

FIG. 4 shows an example of information used for detecting an uncertain portion of the recognition result. The matching distance value of the first place and the matching distance ratio of the second place and the first place of the recognition processing result are shown. It is a frequency distribution. The white outline is the correct answer, and the shaded area is the error case. The reliable correct answer is the ratio of the second place to the first place (relative value) when the matching distance value (insulation value) of the first place is less than 0.1.
It is obtained when is set to 1.4 or more. In other words, when the matching distance value of the first place is 0.1 or more, or the ratio of the second place to the first place is less than 1.4, the recognition result is uncertain, so that the error can be reliably rejected.

In addition, it is considered that the location where there is a possibility of misrecognition in the recognition result is determined based on the error tendency obtained after a large number of recognition experiments (for example, plosive sounds / p, t, k, b, d, syllables including g /).

Furthermore, when the method of the present invention is incorporated after the Kana-Kanji conversion, the conversion uncertainty is detected using a scale for measuring the likelihood of the converted sentence (for example, Japanese). As a scale, the frequency information of independent words, the possibility of combining independent words and adjunct words as bunsetsu, and the determination of the minimum bunsetsu method are considered.

Hereinafter, an embodiment of the present invention will be described in detail with reference to FIG. FIG. 5 is a block diagram of an embodiment of a text input / processing / converting device by voice input according to the present invention. In the input / recognition unit 1, a sentence or the like is input via voice input means such as a microphone 11. The input voice is subjected to voice analysis by the voice recognition unit 12 and then recognized through a matching process with a standard voice. At this time, the uncertainty part of the recognition result is obtained based on the matching distance of the 1st place of the recognition result, the matching distance ratio of the 2nd place and the 1st place, and the like as described above. Since the configuration of the voice recognition unit is not the main object of the present invention, a specific embodiment will be omitted.
It can be achieved using any of the conventional methods. Next, the display of the recognition / sentence conversion result of the input voice and the result of the uncertainty part and the display / output part 2 are displayed.
Displayed on 21. The speaker 22 is used as a means for exchanging humans with the system and outputting the recognition result as voice. The control unit (cpu) 3 is composed of, for example, a microcomputer 8086 manufactured by Intel Corp., and is in charge of controlling the input / recognition unit, the display / output unit, etc. and executing the Japanese language processing / conversion. The external file 4 is composed of, for example, a mini floppy (FD) 41, and is used for storing the created (edited) Japanese sentence information containing kana-kanji and the Japanese information processing / conversion program. The printer 5 is composed of a type printer 51, and outputs the text of the edited result in the type.

〔The invention's effect〕

According to the present invention, the time and effort required for confirmation and correction can be shortened by recognizing mainly the location of an input error and automatically moving the correction cursor, so that efficient text editing is possible.

[Brief description of drawings]

FIG. 1 is a diagram showing a processing flow of a conventional confirmation / correction method, FIG. 2 is a diagram showing a processing flow of a text input method incorporating a display, confirmation, and correction method according to the present invention, and FIG. FIG. 4 is a diagram showing an example of display on the display, FIG. 4 is a diagram showing an example of information used for detecting an uncertain portion of a recognition result, and FIG. 5 is a diagram showing an embodiment of the present invention. Is. 1 ... Input / recognition section, 2 ... Display / output section.

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Akio Amano 1-280 Higashi Koigakubo, Kokubunji, Tokyo Inside Central Research Laboratory, Hitachi, Ltd. (72) Inventor 熹 Ichikawa 1-280, Higashi Koigakubo, Kokubunji, Tokyo Hitachi Ltd. Central Research Laboratory (56) Reference JP-A-56-162142 (JP, A) JP-A-59-99532 (JP, A)

Claims (3)

[Claims] 1. Input means for inputting voice, voice recognition means for analyzing / recognizing the input voice, display means for displaying a result recognized by the voice recognition means, and recognition result for the voice. If it is uncertain, the control for displaying the uncertain portion in the recognition result displayed on the display means in a distinguishable manner and for positioning the correction cursor at the uncertain portion A voice-based text input device, which is composed of 2. The voice recognition means has a predetermined matching distance reference, and makes the voice recognition result uncertain when the voice recognition result indicates a reference value or more. A voice-based text input device according to claim 1. 3. The voice recognition means obtains the matching distance of the input voice, and when the ratio of the first-order matching distance to the second-order matching distance is less than a predetermined reference,
The voice-based text input device according to claim 1, wherein the voice recognition result is uncertain.