JP2708771B2 - Voice recognition method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Regarding a speech recognition system for displaying and evaluating a recognition result obtained by recognizing an input speech, an independent word or a whole phrase that is a candidate for a recognition result is displayed, and a part of the independent word is displayed. For the purpose of prompting the correctness of the correct word and the correctness of the part of the adjunct separately, and quickly evaluating the recognition result, of the recognition results obtained by recognizing the input speech, the clauses of candidates having the highest priority are sequentially selected. Only independent words are displayed, corresponding to the correct answer input by the user, the entire phrase including the independent word is sequentially displayed, and a correct phrase corresponding to the correct answer input by the user is output. In addition, among the recognition results obtained by recognizing the input speech, the phrases of the candidates having the highest priority are displayed in order, and the whole input by the user is correct, only the independent word is correct, and the whole corresponds to the error. Then, the correct phrase is output, the entire phrase including the correct independent word is sequentially displayed, and the next candidate phrase is sequentially displayed.

[Industrial applications]

The present invention relates to a speech recognition method for displaying and evaluating a recognition result obtained by recognizing input speech.

[Problems to be solved by conventional technology and invention]

Conventionally, voice recognition results are displayed on a display in descending order of priority, and a user (user)
To output the correct recognition result.

For this reason, the recognition result candidates are sequentially displayed on the display from the first place, and if the first place candidate has an error, only the auxiliary word having the same independent word of the first place candidate is different. Second and third place phrases are displayed,
There is a problem that a desired phrase cannot be quickly displayed and selected. More specifically, for example, when “boiled” is input as an input voice, and the recognition result is the first candidate: I am the second candidate: I am the third candidate: I am also the fourth candidate: boiled, The first candidate "I am" is displayed on the display at the first time, and if an error is sequentially instructed, the second candidate "I am" at the second time and the third candidate "I am" at the third time Is also displayed. Further, the correct answer "boiled" is displayed for the first time for the first time, and is output by designating this as the correct answer. In this manner, similar phrases are repeatedly displayed in the order of priority, even though the user has once instructed that the phrase is incorrect, and it takes time to display and evaluate a desired phrase.

The present invention displays the independent words or the entirety of a phrase that is a candidate for the recognition result, prompts the correctness of the independent word portion and the correctness of the attached word portion separately, and quickly evaluates the recognition result. The purpose is.

[Means to solve the problem]

 FIG. 1 shows a principle configuration diagram of the present invention.

In FIG. 1, a phrase recognition unit 1 analyzes a speech input.
The matching is performed to generate a recognition result.

The result display unit 2 sequentially displays the independence word of the candidate phrase having the highest priority of the recognition result or the entire phrase, and sequentially displays the phrase including this independence word in response to the user's correct instruction of the independence word. This is to instruct the correct phrase.

[Action]

According to the present invention, as shown in FIG. 1, among the recognition results obtained by analyzing and matching the speech input by the phrase recognition unit 1, the result display unit 2 displays the entire phrase or only the independent word of the candidate with high priority in order. The entire phrase including the self-sustaining word is sequentially displayed in accordance with the user's correct answer instruction for this display, and the correct phrase is output in accordance with the correct answer input by the user.

Therefore, when the correct independent word or a phrase including the correct independent word is displayed on the display, the user indicates the correct independent word as a correct answer, and further, when the phrase including the correct independent word is displayed, the relevant phrase is displayed. When the user instructs the correct answer, the user can quickly select and output a correct phrase with a small number of operations.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be sequentially described in detail with reference to FIGS.

In FIG. 2, the result display section 2 comprises 2.1 to 2.7.

The selection unit 2.1 skips a candidate phrase including an independent word (independent word corresponding to a word number) whose flag stored in the memory 2.2 is “1” in the recognition result input from the phrase recognition unit 1. , The next candidate clause is sent to the switch 2.3. As a result, the display of the candidate phrase including the independent word that the user has indicated not to be the correct answer is skipped, and the number of operations of the user instruction is reduced.

As shown in FIG. 5, the memory 2.2 stores information in which word numbers are associated with flags. In the initial state, all flags are "0" (zero). The flag of the word number corresponding to the independent word that the user has indicated not to be the correct answer is set to “1”.

The switch 2.3 responds to the instruction of the controller 2.4,
The candidate sent from the selection section 2.1 is switched to either the independent word display section 2.5 or the phrase display section 2.6 and sent.

When the user indicates that the displayed independent word is correct, the controller 2.4 switches the switch 2.3 and switches the candidate sent from the selection unit 2.1 to be sent to the phrase display unit 2.6.

The independent word display unit 2.5 displays only an independent word portion of a phrase received as a recognition result. This displayed independent word is stored in the temporary memory 2.7.

The phrase display unit 2.6 displays the entire phrase received as a recognition result.

The temporary memory 2.7 temporarily stores the independent word displayed by the independent word display 2.5. This temporary memory
Independent words stored in 2.7 are stored in memory before the next candidate is displayed if the user indicates that the independent word is not correct.
Sent to 2.2. Then, the flag of the corresponding word number is set to 1.

Next, in the order shown in the flowchart of FIG.
The operation of the configuration in FIG. 2 will be described in detail with reference to FIG. 5 and FIG.

In FIG. 3, a first candidate is input. This means that the first candidate among the recognition results recognized by the phrase recognition unit 1 in FIG. 2, for example, the first candidate “I am” in FIG.
This means inputting to the selection unit 2.1.

In the figure, only independent words are displayed. This is, for example, the independent word “I” of the first candidate “I am” entered in the figure.
Is displayed on the display by the independent word display unit 2.5 in FIG.

In the figure, the user indicates whether or not the answer is correct. this is,
In the figure, this means that the user indicates whether or not the recognition result of the self-sustained word displayed on the display, for example, “I” is correct as the uttered voice. In the case of YES (correct answer), the following is performed in the figure. If NO (if wrong), execute the steps in the figure.

In the figure, independent words are stored. This corresponds to, for example, the user indicating that the first candidate independent word “I” displayed on the display by the independent word display unit 2.5 in FIG. 2.2
As shown in FIG. 5, this means that the flag of the word number i corresponding to the independent word "I" is set to "1".

In the figure, the next candidate is input. This means that, for example, the second candidate "I" in FIG. 4 is input to the selection unit 2.1.

In the figure, it is determined whether the word is a new independent word. this is,
For example, the second candidate "I" is inputted by the selection unit 2.1 in FIG.
Means that there is no new independent word.
In this case, since the flag of the word number i of the independent word "I" in FIG. 5 stored in the memory 2.2 in FIG. 2 is set as "1", it is determined that the word is not a new independent word. Is determined to be NO. Then, similarly, the third candidate “I also”, which is the next candidate in the figure, is input. However, this is also determined to be NO in the figure because it is not a new independent word. Furthermore,
When the fourth candidate “boiled”, which is the next candidate in the figure, is input, this is determined to be YES in the figure because the new independent word is “boiled”, and the independent word “boiled” is displayed in the figure. Then, if the user instructs a correct answer in the figure, the result is YES in the figure, and the following steps are executed.

In the figure, the entire phrase is displayed. This means, for example, that the entire phrase "boiled" of the independent word "boiled" is displayed in the figure.

In the figure, the user indicates whether or not the answer is correct. In the case of YES, the whole of the correct phrase in the figure, for example, "boiled" is output. In the case of NO, the next candidate including the same independent word in the figure is input, and the processing in the figure is repeated.

As described above, the recognition result is displayed sequentially from the first candidate independent word, and when the user instructs that the correct answer is displayed, the entire phrase including the correct independent word is displayed, and the user is further correct. By instructing and outputting a correct phrase, the user can quickly select and output a correct candidate with a small number of instructions (the number of operations).

FIG. 4 shows an example of a speech recognition result. This shows an example of a recognition result obtained by recognizing the input voice “boiled” by the phrase recognition unit 1. The recognition results are prioritized as a first candidate, a second candidate, a third candidate, and a fourth candidate as shown in the figure.

FIG. 5 shows an example of the independent word information held in the memory 2.2. This is because, when the independent word display unit 2.5 displays on the display and indicates that the user is wrong (NO in FIG. 3), the flag of the word number corresponding to the independent word is set to “1”. Is set to. For example, the flag of the word number i corresponding to "I" is set to "1", and it is displayed that the candidate independent word is incorrect.

Next, FIG. 6 shows that the user selects either YES / NO (correct / error) for the entire phrase including the independent word displayed on the display or the correct independent word shown in FIG. In contrast to giving one instruction, this
S1 (correct clause), YES2 (only independent words), NO
One of three (errors) can be given and given.

In FIG. 6, the selection unit 3.1, which has received the recognition result input from the phrase recognition unit 1, sends to the phrase display unit 3.6 a candidate that does not include an independent word designated by the user as an error in the memory 3.2, Display the entire phrase on the display. In response to the display of this entire phrase, the user
Indicate any one of the following three: correct (phrase is correct), YES2 (correct only independent word), NO (error). For example, when NO is instructed, the independent word of the displayed phrase is notified to the memory 3.2 via the temporary memory 3.7, and the flag of the word number corresponding to the independent word stored in the memory 3.2 is set to “ Set to 1 ". In response to this set, when the fetched next candidate includes the independent word whose flag is set to “1”, the selection unit 3.1 skips and fetches the next candidate. On the other hand, if YES2 is specified, the selection unit 3.
1 displays the entire phrase of the candidate including the correct independent word among the next taken in candidates. When YES1 is specified, the correct phrase is output.

As described above, the entire candidate phrase is displayed on the display, and the user is prompted to input an instruction indicating that the entire phrase is correct, only the independent word is correct, and that the entire phrase is incorrect. It is possible to quickly evaluate and output.

〔The invention's effect〕

As described above, according to the present invention, when a correct independent word or a phrase including the correct independent word is displayed on the display, the user indicates the correct independent word as a correct answer, and further includes the correct independent word. When a phrase is displayed, the user designates the phrase as a correct answer and evaluates and outputs the correct phrase, so that the user can quickly evaluate and output the correct phrase with a small number of operations. Can be.

[Brief description of the drawings]

FIG. 1 is a block diagram of the principle of the present invention, FIG. 2 is a block diagram of one embodiment of the present invention, FIG.
FIG. 4 shows an example of a speech recognition result, FIG. 5 shows an example of independent word information held in a memory 2.2, and FIG. 6 shows another embodiment of the present invention. In the figure, 1 is a phrase recognition section, 2 and 3 are result display sections, 2.1 and 3.1.
Is a selection section, 2.2 and 3.2 are memory, 2.3 is a switch, 2.4 and 3.
4 is a controller, 2.5 is an independent word display section, 2.6 and 3.6 are phrase display sections, and 2.7 and 3.7 are temporary memories.

Claims (2)

(57) [Claims] In a speech recognition system for displaying and evaluating a recognition result obtained by recognizing an input voice, only autonomous words are sequentially selected from candidate phrases having a high priority order among recognition results obtained by recognizing an input voice. And displaying the entire phrase including the independent word in response to the correct answer input by the user, and outputting a correct phrase corresponding to the correct answer input by the user. Voice recognition method. 2. A speech recognition method for displaying and evaluating a recognition result obtained by recognizing an input voice, wherein the recognition results obtained by recognizing the input voice are sequentially displayed in the order of a candidate having a higher priority. The whole user input is correct, only the independent word is correct, the whole corresponds to the error, the correct phrase is output, the entire phrase including the correct independent word is sequentially displayed, the next candidate phrase is sequentially displayed A speech recognition method characterized by being configured to be displayed.