Speech Recogniton System, Speech Recognition Method, and Program

• US 20090234648A1
• Filed: 10/11/2006
• Published: 09/17/2009
• Est. Priority Date: 10/13/2005
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Nagatomo, Kentarou
• Granted Patent: US 8,214,209 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Disclosed is a speech recognition system which including speech input means for receiving the speech data, speech recognition means for receiving the input speech data from the speech input means and performing speech recognition, recognition result evaluation means for determining a priority of at least one of a recognition result and each portion forming the recognition result obtained by the speech recognition by the speech recognition means, storage means for storing the recognition result and the priority, recognition result formatting means for determining display/non-display of the recognition result and/or each portion forming the recognition result and generating output information according to the priority, and output means for outputting the output information.

Claims

1. A speech recognition system comprising:

speech input means that inputs speech data;

speech recognition means that receives the speech data input from said speech input means and performs speech recognition;

recognition result evaluation means that derives a confidence of a recognition result obtained by the speech recognition by said speech recognition means and determines a priority of the recognition result based on the confidence;

storage means that stores at least the recognition result and the priority;

recognition result formatting means that performs formatting of output information on the recognition result according to the priority; and

output means that outputs the output information that has undergone formatting by said recognition result formatting means.

2. The speech recognition system according to claim 1, wherein said recognition result formatting means changes whether to output the recognition result or not, and/or an output form of the recognition result according to the priority.

3. The speech recognition system according to claim 1, wherein said recognition result formatting means performs variable control over the output form of the recognition result according to the priority to output the recognition result with the priority of a comparatively high value in the output form capable of being distinguished more easily than the recognition result with the priority of a comparatively low value.

4. The speech recognition system according to claim 3, wherein said recognition result formatting means performs variable control over the output form of the recognition result according to the priority to adjust at least one of an abbreviation form, an output size, grayscale output, a balloon, animation, see-through output, superimposed output, and an output speed of the recognition result.

5. The speech recognition system according to claim 1, wherein said recognition result evaluation means determines the priority of at least one of the recognition result and a portion that forms the recognition result, obtained by the speech recognition by said speech recognition means.

6. The speech recognition system according to claim 1, wherein said recognition result evaluation means derives a confidence of a word obtained by the speech recognition by said speech recognition means, and computes the priority based on the confidence and an importance level of said word.

7. A reproduction device including the speech recognition system according to claim 1, wherein said speech input means inputs the speech data recorded on a recording medium and supplies the received speech data to said speech recognition means, and the recognition result formatted according to the priority of a predetermined level is output from among speeches recorded on said recording medium.

8. A speech transcription device that receives speech data, performs speech recognition, and outputs a recognition result, comprising:

the speech recognition system according to claim 1.

9. The speech recognition system according to claim 1, wherein

said recognition result evaluation means determines the priority of the recognition result and/or a portion that forms the recognition result, obtained by the speech recognition by said speech recognition means;

said recognition result formatting means determines display/non-display of the recognition result and/or the portion that forms the recognition result and outputs an output image, according to the priority; and

said output means includes screen displaying means for outputting the output image to a display device.

10. The speech recognition system according to claim 9, further comprising:

display target range setting means that sets a range of the speech data to be output and a range of at least one of the recognition result and the portion of the recognition result;

said recognition result formatting means determining display/non-display of the recognition result and/or the portion of the recognition result and creating the output image according to said priority and set range of said output.

11. The speech recognition system according to claim 9, wherein the input speech data supplied to said speech recognition means from said speech input means is stored in said storage means;

and

the speech recognition system further include;

speech reproduction instructing means that refers to the output image and specifying reproduction of the input speech data; and

speech reproduction means that reproduces the input speech data stored in said storage means, based on the specification by said speech reproduction instructing means.

12. The speech recognition system according to claim 1, wherein said recognition result evaluation means derives the priority using at least one of an acoustic likelihood, a linguistic likelihood, and a confidence based on a posterior probability of at least one of the recognition result and the portion of the recognition result obtained by said speech recognition means.

13. The speech recognition system according to claim 1, wherein said recognition result evaluation means derives the priority of a word or a chain of words that may form the recognition result and/or the portion of the recognition result obtained by said speech recognition means, using an importance level.

14. The speech recognition system according to claim 9, wherein, when the output image is created, said recognition result formatting means displays at least one of the recognition result and the portion of the recognition result with the priority thereof exceeding a predetermined threshold value.

15. The speech recognition system according to claim 9, wherein, when the output image is created, said recognition result formatting means performs sequential display of the recognition result and/or the portion of the recognition result in descending order of the priority, and when there is not an enough display region of said display device, said recognition result formatting means performs alternative display of a remainder of the recognition result and/or the portion of the recognition result, using a predetermined symbol.

16. The speech recognition system according to claim 9, wherein, when the output image is created, said recognition result formatting means performs control so that an area of the display region occupied by the recognition result and/or the portion of the recognition result increases in ascending order of the priority, and a larger area of the display region is occupied by the recognition result and/or the portion of the recognition result with the priority of a relatively high level, and a smaller area of the display region is occupied by the recognition result and/or the portion of the recognition result with the priority of a relatively low level.

23. A server device including the speech recognition system according to claim 1.

24. A portable communication terminal device including the speech recognition system according to claim 1.

17. A method of performing speech recognition by a speech recognition device, said method comprising:

inputting speech data;

performing speech recognition processing on the input speech data;

deriving a confidence of at least one of a recognition result and a portion of the recognition result obtained by the speech recognition processing and determining a priority of the at least one of the recognition result and the portion of the recognition result based on the confidence;

a step of performing formatting of output information on the recognition result and/or the portion of the recognition result according to the priority; and

a step of outputting the formatted output information.

18. The method of performing speech recognition according to claim 17, further comprising:

determining display/non-display of the recognition result and/or the portion of the recognition result and creating an output image according to the priority; and

outputting the created output image onto a display screen.

19. The method of performing speech recognition according to claim 18, further comprising:

setting a range of the speech data to be output and a range of the at least one of the recognition result and the portion of the recognition result;

in creating the output image, the display/non-display of the recognition result and/or the portion of the recognition result being determined and the output image being created according to priority and set range of said output.

20. The speech recognition method according to claim 18, further comprising:

referring to the output image and reproducing the input speech data; and

reproducing the speech data based on the specification.

21. A program that causes a computer which performs speech recognition processing to execute:

a processing of inputting speech data;

a processing of performing speech recognition on the input speech data;

a processing of deriving a confidence of at least one of a recognition result and a portion of the recognition result obtained by the speech recognition and determining a priority of the at least one of the recognition result and the portion of the recognition result based on the confidence;

a processing of performing formatting of output information on the recognition result and/or the portion of the recognition result according to the priority; and

a processing of outputting the formatted output information from an output device.

22. The program according to claim 21, which causes the computer to execute:

a processing of setting a range of the speech data to be output and a range of the at least one of the recognition result and the portion of the recognition result; and

a processing of determining display/non-display of the recognition result and/or the portion forming the recognition result and creating an output image according to said priority and set range of said output.

25. A speech recognition system comprising:

means that derives confidences of speech recognition results of speech data input in a time series, computes priorities of the recognition results based on the confidences, and stores the recognition results and the priorities in storage means, with the recognition results associated with the priorities; and

control means that performs control over output of the recognition results based on the priorities of the recognition results;

said control means performing the control so that, from among the recognition results of a plurality of speeches included in the input speech data being comparatively long, a portion with the priority of a relatively high level is selectively output, or the portion with the priority of the relatively high level is output in an output form different from an output form of a portion with the priority level of a relatively low level, to allow the recognition results of the speeches to be overlooked.