Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1998-09-03
2001-07-10
Tsang, Fan (Department: 2748)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S276000
Reexamination Certificate
active
06260015
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to the field of correcting errors in natural speech recognition applications, and in particular, to correcting natural speech recognition errors in languages in which words and phrases are represented by characters instead of groups of letters.
2. Description of Related Art
Many languages, such as Chinese and other Asian languages, are based on written characters to represent words or phrases, rather than groups of letters. It is very difficult and time consuming to manually enter the characters of such languages into a computer. Even knowing that each character represents a word or phrase in its entirety, it is often shocking to learn, for example, that there are over 7,000 Chinese characters. In order to use a standard keyboard with less than 100 keys, operators must go through a great deal of practice and training.
Speech recognition is an extremely fast method of entering Chinese characters. It is much faster than typing or using handwriting recognition.
Unfortunately, speech recognition in any language is not 100% accurate. Mistakes are made that must be corrected. Moreover, unless the mistakes are corrected according to the rules of the speech application, the speech application will not be self-correcting for future dictation. Correction is that much more difficult in character based languages. If a user is forced to correct speech recognition errors using a standard keyboard the effectiveness of speech recognition is reduced because the user still has to learn how to enter Chinese characters, or the characters of other languages, using a keyboard and/or mouse.
Presently, in Chinese and other Asian languages, it is difficult to correct speech recognition errors with a keyboard and mouse. This method describes a natural, user interface that will allow users to easily and naturally use speech recognition with a standard computer without purchasing an extra equipment or going through a great deal of training.
There is a long felt and unsatisfied need to enable efficient and convenient correction of characters with a standard keyboard and mouse, that is, by standard cursor controls.
SUMMARY OF THE INVENTION
In accordance with the inventive arrangements, a graphical user interface (GUI) allows a user to correct speech recognition errors naturally, and does not require extensive training. The GUI reflects the complementary technologies of speech and handwriting recognition technology. Characters can be “written” into a portion of the GUI, in whole or in part, by the standard cursor controls provided by a standard keyboard and mouse.
A method for correcting errors of a speech recognition application in natural speech recognition of words or phrases represented by characters, in accordance with an inventive arrangements, comprises the steps of: selecting a character corresponding to an incorrectly recognized spoken word or phrase; observing a list of alternative characters; replacing the incorrect character with one of the alternative characters, if the one of the alternative characters correctly represents the incorrectly recognized word or phrase; and, drawing at least some features of a character correctly representing the incorrectly recognized word or phrase by moving a cursor, if none of the alternative characters in the list correctly represents the spoken word or phrase.
The method can further comprise the step of updating the list of alternative characters responsive to the drawing step.
The updating step can comprise the steps of: filtering a comprehensive list of the characters based on the drawing step; regenerating the list of alternative characters based on the filtering step; and, repeating the observing and replacing steps. As necessary, the method further comprises the steps of: repeating the drawing step if none of the alternative characters in the regenerated list correctly represents the spoken word or phrase; and, repeating the filtering, regenerating, observing and replacing steps.
Advantageously, the drawing step can comprise the steps of: selecting at least one predrawn character feature from a list of the features; and, pasting the at least one selected feature into at least one of a plurality of predefined zones by which the characters are defined.
The method can further comprise the step of updating the list of alternative characters responsive to the selecting and pasting steps.
The updating step can comprise the steps of: filtering a comprehensive list of the characters based on the selecting and pasting steps; regenerating the list of alternative characters based on the filtering step; and, repeating the observing and replacing steps. As necessary, the method can further comprise the steps of: repeating the selecting and pasting steps if none of the alternative characters in the regenerated list correctly represents the spoken word or phrase; and, repeating the filtering, regenerating, observing and replacing steps.
The method can also further comprise the steps of: displaying the list in a first part of a graphical user interface (GUI) of the speech recognition application; and, drawing the at least some features in a second part of the GUI. The appearance of the cursor can be changed at least within the second part of the GUI. The updating step can be initiated automatically when the cursor is moved out of the second part of the GUI.
A method for correcting errors of a speech recognition application in natural speech recognition of words or phrases represented by characters, in accordance with another inventive arrangement, comprises the steps of: selecting a character corresponding to an incorrectly recognized spoken word or phrase; observing a list of alternative characters; replacing the incorrect character with one of the alternative characters, if the one of the alternative characters correctly represents the incorrectly recognized word or phrase; drawing at least some features of a character correctly representing the incorrectly recognized word or phrase by moving a cursor, if none of the alternative characters in the list correctly represents the spoken word or phrase; updating the list of alternative characters responsive to the drawing step; and, repeating the replacing, drawing and updating steps until a character correctly representing the word or phrase is selected from the updated list
The updating step can comprise the steps of: filtering a comprehensive list of the characters based on the drawing step; and, regenerating the list of alternative characters based on the filtering step.
The drawing step can advantageously comprise the steps of: selecting at least one predrawn character feature from a list of the features; and, pasting the at least one selected feature into at least one of a plurality of predefined zones by which the characters are defined.
The method can further comprise the steps of: displaying the list in a first part of a graphical user interface (GUI) of the speech recognition application; selecting the at least one predrawn character from a second part of the GUI; and, pasting the at least one feature into a third part of the GUI. The appearance of the cursor can be changed at least within the third part of the GUI. The updating step can initiated automatically when the cursor is moved out of the third part of the GUI.
A graphical user interface (GUI), embodied in a programmed routine set of instructions for correcting errors of a speech o recognition application in natural speech recognition of words or phrases represented by characters, in accordance with yet another inventive arrangement, comprises: a first section for displaying a list of alternative characters for replacing incorrect characters; a second section for drawing at least some features of a character correctly representing said incorrectly recognized word or phrase by moving a cursor; means for providing a cursor with different appearances in said first and second sections respectively; and, means for initiating regeneration and redisplay of said list responsive to infor
Gong Oing
Lewis James R.
VanBuskirk Ronald
Wang Huifang
Akerman & Senterfitt
International Business Machines Corp.
Opsasnick Michael N.
Tsang Fan
LandOfFree
Method and interface for correcting speech recognition... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and interface for correcting speech recognition..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and interface for correcting speech recognition... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2466003