JP2003099091A - Speech recognition device and speech recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech recognition device in which user's reuttering is not required, work load of the user is made light and quick and precise correcting and supplementing functions for recognition results are provided. SOLUTION: A search processing section 3 conducts a recognition process for voice recorded in a voice storage section 1 using grammar information and a recognition dictionary section 2. A search control section 4 transmits recognition result candidate word strings obtained by the recognition process of the section 3 to a recognition result candidate display section 5 and makes a request to the section 3 so that the section 3 conducts voice recognition employing only a candidate word string, that starts with a correct character string notified by a correct character storage section 8, as a recognition object word string and voice recorded in a voice storage section. A correct character input section 7 inputs the characters of the correct word string one character at a time. The section 8 records the correct character string inputted from the section 7 when the correct word string is not included in the recognition result candidate word string displayed in the section 5 and notifies the above to the section 4.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device and a voice recognition method, and more particularly to a voice recognition device and a voice recognition method capable of correcting an erroneous recognition result quickly and highly accurately.

[0002]

2. Description of the Related Art An example of a method for correcting an erroneously recognized word in a conventional voice recognition device is "Personal computer moves with one's own voice (by Hiroko Murakami, published on August 10, 1999 by NEC Creative Co., Ltd.), 22-25. It is listed on the page. In this conventional technique, as a correction of a word that has been erroneously recognized, "a plurality of recognition result candidates are listed and a user selects one of them" and "A rerecognition process and a re-recognition process are performed to replace the erroneous recognition result with the correct word. "," Correct by inputting the correct answer result of the erroneously recognized part with the keyboard ".

On the other hand, in the character input device and the character recognition device, a "complementary function" has been invented in which the character string input device side complements the character string even if the user does not type in all the desired character string. The conventional example is "http: //www.hll
a.is.tsukuba.ac.jp/~yas/ipe/nitiniti2-enshu-1996/1
996-10-21 / unix-completion.html ".
This is to display and select only the correct answer candidate character string corresponding to the character string input by the user, and when the candidates are few or uniquely decided during the character string input, the entire character string is not input. However, it is a function that can select the correct answer candidate.

[0004]

The above-mentioned conventional speech recognition apparatus, when there is no correct answer in the recognition result candidates, "repeat voice", "input the correct answer on the keyboard, and then perform Kanji conversion", etc. The burden on the user was heavy. Further, in the complementary function in the conventional character input device, etc., although it is not necessary to enter all correct answer character strings, if there are many candidates, for example, if there are hundreds or more similar words such as voice recognition, many still remain. I had to enter the string.

An object of the present invention is to use the correct character string input by the user while avoiding the above-mentioned problems of the prior art,
The recognition dictionary including the backup dictionary is used to re-recognize the erroneously recognized voice data uttered by the user, thereby eliminating the need for the user to re-voice. An object of the present invention is to provide a voice recognition device having a highly accurate recognition result correction function and a complementary function.

[0006]

A first invention of the present application is, in a voice recognition device, a voice storage section for recording an input voice, a recognition dictionary section having recognition word dictionary information, and a prepared grammar. A search processing unit that performs recognition processing on the voice recorded in the voice storage unit using the information and the recognition dictionary unit, and a recognition result candidate word string by the recognition processing of the search processing unit on the recognition result candidate display unit. When the correct answer character string is notified from the correct answer character storage unit, only the candidate word string starting with the notified correct answer character string is recognized as the recognition target word string by using the voice recorded in the voice storage unit. A search control unit that requests re-recognition processing for recognition to the search processing unit, a recognition result candidate display unit that displays the recognition result candidate word string, and a character that inputs characters of the correct word string one by one. Record the correct answer character string input from the correct answer character input section when a correct answer word string is not included in the recognition result candidate word string displayed in the character input section and the recognition result candidate display section. The correct answer character storage unit for notifying the search control unit is also included.

According to a second invention of the present application, the search processing section of the first invention performs the voice recognition by using a backup dictionary having more words than the recognition dictionary section during the re-recognition processing. Is characterized by.

A third invention of the present application is the search control section which outputs an unknown word word registration request upon receipt of the character input end notification in the correct character input section of the first invention, and the unknown word word registration. An unknown word word registration unit for registering a word of the character string recorded in the correct character storage unit in the recognition dictionary unit by an unknown word word input in response to a request. To do.

In a fourth aspect of the present invention, in the voice recognition method, a voice storage section for recording an input voice and a recognition dictionary section having recognition word dictionary information are provided in advance, and the prepared grammar information and the recognition are provided. A recognition process is performed on the voice recorded in the voice storage unit by using the dictionary unit, and a recognition result candidate word string that is a recognition result by the recognition process is displayed on a recognition result candidate display unit that is provided in advance. When the characters of the correct answer word string are input character by character from the correct answer character input section prepared in advance when the correct answer word string is not included in the recognition result candidate word strings displayed in the result candidate display section. , The input correct answer character string is recorded in a prepared correct answer character storage unit, and only candidate word strings starting with the correct answer character string in the recorded correct answer character storage unit are used as recognition target word strings. And performing re-recognition processing for speech recognition using the voice recorded in the voice storage unit.

A fifth aspect of the present invention is characterized in that during the re-recognition process according to the fourth aspect, the voice recognition is performed using a backup dictionary having more words than the recognition dictionary section.

According to a sixth aspect of the present invention, an unknown word word registration request is output in response to the character input end notification in the correct character input section of the fourth aspect, and the unknown word word registration request is input. The registered unknown word is registered in the recognition dictionary unit for the character string recorded in the correct character storage unit.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of a voice recognition apparatus showing an embodiment of the present invention.

Referring to FIG. 1, the voice recognition device of the present invention includes a voice storage unit 1, a recognition dictionary unit 2, and a search processing unit 3.
, Search control unit 4, recognition result candidate display unit 5, candidate selection input unit 6, correct answer character input unit 7, correct answer character storage unit 8, unknown word word registration unit 9, and backup dictionary unit 10.
Composed of and.

The voice storage unit 1 records the input voice.

The recognition dictionary unit 2 has recognition word dictionary information.

The search processing unit 3 uses the grammatical information such as natural language grammar, statistical language model or network grammar, and the recognition dictionary unit 2 to perform recognition processing on the voice recorded in the voice storage unit 1.

The search control unit 4 requests the search processing unit 3 to perform recognition processing of the input voice, passes the recognition result candidate word string to the recognition result candidate display unit 5, and at the same time stores the correct character from the correct character storage unit 8. When a column is notified, only the candidate word string starting with the notified correct character string is displayed in the recognition result candidate display part, and even if all candidate word strings are displayed in the recognition result candidate display part, the recognition result candidate display is still displayed. When there is a space in the copy or when the number of candidate word strings is less than a certain number, only the word string starting with the correct answer character string notified from the correct answer character storage unit 8 is set as the recognition target word string, and the recognition dictionary. As an instruction, the search processing unit 3 is instructed to perform voice recognition again on the voice recorded in the voice storage unit 1 while using the recognition dictionary recorded in the backup dictionary unit 10, and all correct character strings are input. And outputs an instruction of the unknown word word registered in the user receives a notification of completion.

The recognition result candidate display section 5 notifies the user of the candidate word string of the recognition result.

The candidate selection input unit 6 is a recognition result candidate display unit 5
If there is a correct answer in the candidate word string displayed in, the correct candidate word notified by the user is determined.

The correct character input unit 7 is a recognition result candidate display unit 5
When the correct answer word string is not included in, the user inputs characters of the correct answer word string one by one.

The correct character storage unit 8 additionally records the correct character string input from the correct character input unit 7, and notifies the search control unit 4 of the correct character strings input so far.

The unknown word registration unit 9 receives a word registration request from the user and performs word registration of the character string recorded in the correct character storage unit 8.

The backup dictionary unit 10 has not only the recognized word dictionary information in the recognition dictionary unit 2 but also more recognized word dictionary information such as proper nouns and technical terms.

Next, the operation of the speech recognition apparatus of the present invention will be described with reference to FIG. (Step 1): The input voice is recorded in the voice storage unit 1. (Step 2): The search control unit 4 requests the search processing unit 3 to perform recognition processing of the input voice. The search processing unit 3 uses the recognition dictionary unit 2 having grammatical information such as natural language grammar, statistical language model or network grammar, and recognized word dictionary information to recognize the voice recorded in the voice storage unit 1. I do. (Step 3): When the recognition processing is completed, the search processing unit 3 passes the candidate word string of the recognition result to the search control unit 4. The search control unit 4 passes the recognition result candidate word string to the recognition result candidate display unit 5, and notifies the user of the recognition result candidate word string. (Step 4): If there is a correct answer in the candidate word string displayed in the recognition result candidate display unit 5, go to Step 5 and
If there is no correct answer in the candidate word string displayed on the recognition result candidate display section 5, the process goes to step 7. (Step 5), (Step 6): The user can notify the candidate selection / input unit 6 of the number of the correct answer and determine the word.

When the first candidate is the correct answer, the next utterance is performed as it is, so that the first candidate is the correct answer in the search control unit 4 without inputting anything in the candidate selection input unit. Then, the next voice processing can be performed. If either of the above two cases applies, the recognition processing of the voice recorded in the voice storage unit 1 ends. (Step 7): If the user has not entered all the character strings in the correct answer character input section, go to Step 8 and notify that the user has entered all the character strings in the correct answer character input section. Run. (Step 8): The user inputs the correct word string into the correct character input unit 7 character by character.

However, when there is a word string candidate in which the correct answer character string is included in the recognition result candidate display portion 5, the user designates up to what is the correct answer, so that the plural characters from the beginning to the middle You can also input all at once. (Step 9): Correct answer character input unit 7 is correct answer character storage unit 8
In addition, the correct answer character string is additionally recorded. The correct character storage unit 8 is
The search control unit 4 is notified of the correct answer character string input so far. (Step 10): The search control unit 4 displays only the candidate word string starting with the notified correct character string on the recognition result candidate display unit 5. (Step 11): When all the candidate word strings are displayed on the recognition result candidate display unit 5, but there is still space in the recognition result candidate display unit, or when the number of candidate word strings is below a certain number. For example, step 12 is executed. If not, return to step 8. (Step 12): The search control unit 4 makes the search processing unit 3
On the other hand, only the word string starting with the correct character string notified from the correct character storage unit is set as the recognition target word string, and the voice recorded in the voice storage unit 1 is instructed to perform the voice recognition again. At this time, the backup dictionary unit 10 is used as a recognition dictionary.
By using the recognition dictionary recorded in, the recognition processing is performed for more vocabulary than in the first recognition processing. Therefore, when the user utters a word registered in the backup dictionary 10, it is possible to correct the candidate with high accuracy and speed. A candidate word string of a new recognition result is obtained by the re-recognition process, and the process returns to step 3. (Step 13) to (Step 15): Search control unit 4
Judges that the word string corresponding to the character string input by the user cannot be obtained as the recognition result because the user has input an unknown word that is not registered in the recognition dictionary,
The user is notified to perform word registration, and the unknown word registration unit 9 registers the unknown word of the character string recorded in the correct character storage unit 8.

[0028]

As described above, according to the present invention, the correct character string input by the user is used, and the recognition dictionary including the backup dictionary is used for the erroneously recognized voice data uttered by the user. By performing the re-recognition process, it is possible to provide a voice recognition device that does not require a re-voice of the user, has a light burden on the user, and has a quick and highly accurate recognition result correction function and a complementary function. There is.

[Brief description of drawings]

FIG. 1 is a block diagram of a voice recognition device showing an embodiment of the present invention.

FIG. 2 is an operation flow diagram of the voice recognition device of the present invention shown in FIG.

[Explanation of symbols]

1 Voice memory 2 Recognition dictionary section 3 Search processing unit 4 Search control section 5 Recognition result candidate display 6 Candidate selection input section 7 Correct answer input section 8 Correct answer character storage 9 Unknown word registration section 10 Backup dictionary section

Claims (6)

[Claims] 1. In a voice recognition device, a voice storage unit for recording an input voice, a recognition dictionary unit having recognition word dictionary information, and a voice storage unit using previously prepared grammatical information and the recognition dictionary unit. A search processing unit that performs recognition processing on the voice recorded in the unit, and a recognition result candidate word string by the recognition processing of the search processing unit is passed to the recognition result candidate display unit, and the correct character string is stored from the correct character storage unit. When notified, the search processing unit is subjected to re-recognition processing for performing voice recognition using only the notified candidate word string starting with the correct character string as the recognition target word string as the recognition target word string using the voice recorded in the voice storage unit. A requesting search control unit, a recognition result candidate display unit for displaying the recognition result candidate word string, a correct character input unit for inputting characters of the correct answer word string one by one, and the recognition result candidate display If the correct answer word string is not included in the recognition result candidate word strings displayed in the section, the correct answer character string input from the correct answer character input unit is recorded, and the correct answer is notified to the search control unit. A voice recognition device comprising a character storage unit. 2. The voice recognition according to claim 1, wherein the search processing unit performs the voice recognition using a backup dictionary having more words than the recognition dictionary unit during the re-recognition process. apparatus. 3. The search control unit which outputs an unknown word word registration request upon receipt of the character input end notification in the correct character input unit, and an unknown word word input in response to the unknown word word registration request. The voice recognition device according to claim 1, further comprising an unknown word registration unit that registers a word of the character string recorded in the correct character storage unit into the recognition dictionary unit. 4. A voice recognition method, comprising a voice storage unit for recording an input voice and a recognition dictionary unit having recognition word dictionary information in advance, and using the prepared grammatical information and the recognition dictionary unit, A recognition process is performed on the voice recorded in the voice storage unit, a recognition result candidate word string that is a recognition result of the recognition process is displayed on a recognition result candidate display unit that is provided in advance, and is displayed on the recognition result candidate display unit. When the correct answer word string is not included in the recognition result candidate word strings that have been input and the characters of the correct answer word string are input one by one from the prepared correct answer character input unit, the input correct answer is input. A character string is recorded in a correct character storage unit prepared in advance, and only candidate word strings starting with the correct character string in the recorded correct character character storage unit are recorded in the voice memory unit as recognition target word strings. A voice recognition method characterized by performing a re-recognition process for performing voice recognition using the voice that has been recorded. 5. The voice recognition method according to claim 4, wherein during the re-recognition process, the voice recognition is performed by using a backup dictionary having more words than the recognition dictionary unit. 6. The unknown character word registration request is output in response to the notification of the character input completion in the correct character input unit, and the unknown word word input in response to the unknown word registration request is stored in the correct character storage unit. 5. The voice recognition method according to claim 4, wherein the word is registered in the recognition dictionary unit for the character string recorded in.