JPS62150294A - Recognition results checking system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] (Industrial application field) The present invention is a word processor that allows voice input.

Used in data input devices, etc., and allows the user to easily and quickly select the desired recognition result when there are multiple candidates for recognition results obtained by recognizing input speech. , relates to a recognition result confirmation method for voice input.

(Conventional technology) In word processors, data input devices, etc.

If text or data can be input by voice, there will be no need to use a large number of keys on a keyboard, etc., and operation will be extremely easy, especially for beginners.

However, with current technology, it is very difficult to recognize speech and obtain the only recognition result desired by the user.

On the other hand, when a plurality of candidates are obtained as recognition results, there is a very high probability that the desired recognition result will be included.

Therefore, it is recognized that these multiple candidates are sequentially presented according to the user's next candidate presentation instructions until the recognition result desired by the user appears, and the user selects the confirmation result by confirming the results. A result confirmation method has traditionally been adopted.

(Problem to be Solved by the Invention) Among the multiple candidates in the recognition results obtained for a certain input voice, there are candidates that have the same sequence of sounds but are written differently in kanji, etc., i.e. homophone and allograph candidates. is often present. At this time, if the sequence of sounds of these homophone and allograph candidates differs from the sequence of sounds desired by the user, all of these homophone and allograph candidates may be rejected as inappropriate recognition results.

In other words, for the user, when one of multiple homophone and allograph candidates is presented and the user realizes that the sequence of sounds is not the desired arrangement, it is determined that all of the homophone and allograph candidates do not provide the desired recognition result. can. however,
In the conventional recognition result confirmation method, each of the homophone and allograph candidates was also presented sequentially in accordance with instructions from a single next candidate presentation instruction means. For this reason, when multiple homophone and allograph candidates with a different sound sequence from the recognition result desired by the user are selected as recognition result candidates, the user can select the next candidate for each of these homophone and allograph candidates. It was necessary to give instructions for presentation, which caused unnecessary effort.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to eliminate the drawbacks of the prior art and, when a homophone and allograph candidate has a different sound sequence from the desired recognition result, simply respond to an instruction to present the next candidate of li. To provide a recognition result confirmation method capable of rejecting all recognition results and easily and quickly selecting a desired recognition result.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the invention of the present application provides means for recognizing a voice input signal and selecting a plurality of candidates for recognition results of the input signal. , a recognition result confirmation method that sequentially presents next candidates according to a user's next candidate presentation instruction until a recognition result desired by the user is presented, the recognition result confirmation method including at least first and second next candidate presentation instruction means. and when the first next candidate presentation instructing means instructs to present the next candidate, the next candidate which has a different arrangement of sounds or notation from the candidate currently being presented is presented next. ,
The present invention is characterized in that when the second next candidate presentation instructing means issues a next candidate presentation instruction, a next candidate having at least a different sequence of sounds from the candidate currently being presented is presented.

(Operation) In such a recognition result confirmation method of the present invention, the first
Separately from the next candidate presentation instruction means, by providing a second next candidate presentation instruction means for instructing the presentation of a next candidate with at least a different sound sequence, it is possible to ensure that the presented candidates match the recognition result desired by the user. is one of the homophone and allograph candidates having different sound sequences, all of the homophone and allograph candidates are rejected from the recognition results by the instruction by the second candidate presentation instruction means. On the other hand, if the presented candidate is a homophone and allograph candidate with the desired recognition result, that is, a candidate with the same sound sequence but with the notation a, the first next candidate presentation instructing means presents the candidate at that time. The candidate homophone and allograph candidate is presented as the next candidate. This eliminates the need for the user to instruct the presentation of the next candidate for each of the homophone and allograph candidates, making it possible to easily and quickly obtain the desired recognition result.

(Example) Hereinafter, the present invention will be described in detail by giving examples of the present invention and referring to the drawings of the examples.

FIGS. 1 and 2 are configuration diagrams showing mutually different embodiments of the present invention. Voice input into the microphone 101 is recognized by the voice recognition unit 102 or 201. As the speech recognition unit 102° 201, a word speech recognition device, a monosyllabic speech recognition device, etc. can be used. In the 11A embodiment, the speech recognition unit 102 generates all recognition result candidates for one input speech at once, and stores them in the candidate storage unit 103. Further, the user gives instructions to the candidate confirmation section 104 by voice.

In the embodiment shown in FIG.
According to the signal from 02, recognition result candidates for one input voice are sequentially sent to the KM supplementary confirmation unit 202 one by one. Further, the user gives instructions to the candidate confirmation section 202 using the correct answer key 203, the next candidate key 204, and the missing sound key 205.

The operation of the embodiment shown in FIG. 1 will be explained. The speech recognition unit 102 stores all recognition result candidates for one input speech in the candidate storage unit 1.
03 and then switches to command input mode. In the command input mode, 191J is "High", "
It recognizes only three words: "Iie" and "Tsugi." In this way, if three words that are not similar are recognized, error-free recognition is possible.

Candidate confirmation section 104 retrieves the first recognition result candidate from candidate storage section 103 and presents it on display section 105 . At this time, the display section 105 displays “If the answer is correct, “Yes”.

If it's a mistake, say "No," and if the order of the notes is different, say "Tsugi."
Please say this out loud. ” is displayed at the same time.

After confirming the presented recognition result candidates, the user follows the instruction message and selects "Yes" if the candidate is the desired recognition result, and "No" if the candidate is not the desired recognition result.
, if the sequence of sounds is different from the desired recognition result, then it is "Tsugi"
The command is pronounced into the microphone 101.

The speech recognition unit 102 recognizes this command and sends the result to the candidate confirmation unit 104.

If the command is "high", the candidate confirmation section 104 outputs the recognition result candidate presented on the display section 105 to the data processing section 106, and then switches the speech recognition section 102 to the next speech input mode.

On the other hand, if the command from the user is "yes", the candidate confirmation section 104 retrieves the next recognition result candidate from the candidate storage section 103. The candidate to be retrieved may be a homophone or allograph from the displayed candidate, or it may be a candidate with a different sequence of sounds. The retrieved candidates are displayed on the display section 10 as well.
5 and waits for command input from the user. Also,
If the command from the user is "Tsugi", candidate confirmation part 1
04 deletes from the candidate storage unit 103 all candidates that have the same sound order as the displayed candidate, and then deletes them from the candidate storage unit 1.
The next recognition result candidate is extracted from 03 and presented on the display unit 105, and waits for a command input from the user.

Therefore, when the user determines that the sound arrangement of the candidate presented on the display unit 105 is different from the desired recognition result, the user can change the candidate to a homophone and allograph by inputting the command "tsugi". There is no need to recognize it.

The data processing unit 106 performs processing such as word processing using the received recognition results.

Next, the operation of the embodiment shown in FIG. 2 will be explained. Speech recognition unit 2
01, when one recognition result candidate is obtained for one input voice, the recognition result candidate is sent to the candidate confirmation unit 202. At this time, the speech recognition unit 201 interrupts the recognition process and waits for an instruction signal from the candidate confirmation unit 202.

After presenting this candidate on the display section 105, the candidate confirmation section 202 waits for an instruction from the user.

The user checks the recognition result candidates presented on the display unit 105, and if the candidate is the desired recognition result, presses the correct answer key 203, and if the candidate is not the desired recognition result, presses the next candidate key 204.
If the sequence of sounds is different from the desired recognition result, select the next note key 2.
Press 05 respectively.

When the correct answer key 203 is pressed, the candidate confirmation unit 202 outputs the recognition result candidates presented on the display unit 105 to the data processing unit 106, and then switches the state of the voice recognition unit 201 to wait for input of the next voice. . On the other hand, if the next candidate key 204 is pressed, the candidate confirmation section 202 sends a signal to the speech recognition section 201 to output the next recognition result candidate. The speech recognition unit 201 resumes the interrupted recognition process, and when it obtains the next candidate, sends it to the candidate confirmation unit 202 and interrupts the recognition process again.

The candidates sent to the candidate confirmation unit 202 may be homophones or have different sound sequences from the displayed candidates. The sent candidates are also displayed on the display section 105.
and wait for instructions from the user. On the other hand, primary tone key 2
If 05 is pressed.

Candidate confirmation unit 202 sends an instruction signal to voice recognition unit 201 to output a candidate whose sound order is different from the currently displayed candidate. Normally, the speech recognition unit 201 sequentially obtains and outputs a plurality of candidates with different notations for one sound sequence. Therefore, when the above-mentioned instruction signal is received, the processing for the sound arrangement at that time is stopped, and a candidate with a different sound arrangement is obtained and sent to the candidate confirmation section 202. The candidate confirmation unit 202 presents the accepted candidates on the display unit 105 and waits for instructions from the user.

Therefore, when the user determines that the sound sequence of the candidate presented on the display unit 105 is different from the desired recognition result, the user can press the public key 205 to respond to the candidate and the homophone and allograph candidate by pressing the public key 205. There is no need to confirm.

One embodiment of the present invention has been described above, but in the embodiment shown in FIG. The process may be interrupted and wait for an instruction from the candidate confirmation unit 202 when the process is obtained. If you do this,
Since the recognition process can be performed in parallel while the user is checking the candidates, it is possible to quickly present the next candidate.

In addition, if the speech recognition unit 201 does not necessarily return recognition result candidates with the same sound arrangement consecutively, when the public key 205 is pressed, the sound arrangement of the candidates presented at that time is suddenly stored. After that, if you get a candidate with the same sequence of sounds, you can immediately reject that candidate.

Furthermore, the method of presenting recognition result candidates, the method of instructing the presentation of next candidates, etc. are not limited to the methods in the above embodiments.

(Effects of the Invention) As explained above, according to the present invention, even if a plurality of homophone and allograph candidates with a different sound sequence from the desired recognition result are obtained, a single instruction to present the next candidate can be used. It is possible to provide a recognition result confirmation method that can reject all homophone and allograph candidates from the recognition results and easily and quickly select a desired recognition result.

[Brief explanation of drawings]

FIG. 1 and FIG. 2 are configuration diagrams showing mutually different embodiments of the present invention. 101...Microphone, 102...Speech recognition section, 103...Candidate storage section, 104...Candidate confirmation section,
105...Display section, 106...Data processing section, 20
1... Speech recognition unit, 202... Candidate confirmation unit, 203
...Correct answer key. 204...Next candidate key, 205...Public key. Agent Patent Attorney Shinsuke Honjo Figure 1

Claims (1)

[Claims] A plurality of candidates for the recognition result of the input signal obtained by recognizing the voice input signal are sequentially selected according to the next candidate presented by the user until the recognition result desired by the user is presented. In the presented recognition result confirmation method, at least the first
and a second next candidate presentation instructing means, and when the first next candidate presentation instructing means instructs to present the next candidate, the arrangement of sounds or the way of notation is different from the candidate currently being presented. A next candidate is presented next, and when a next candidate presentation instruction is given by the second next candidate presentation instruction means, a next candidate whose sound sequence is at least different from the candidate currently being presented is presented. recognition result confirmation method.