Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-09-27
2002-07-09
Dorvil, Richemond (Department: 2748)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S257000, C704S235000, C704S231000
Reexamination Certificate
active
06418410
ABSTRACT:
CROSS-REFERENCE TO RELATED APPLICATIONS
(Not Applicable)
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
(Not Applicable)
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates generally to speech dictation systems, and more particularly to a method of updating language models in speech recognition engines of speech applications during sessions in which speech misrecognitions are corrected.
2. Description of Related Art
Speech recognition is the process by which an acoustic signal received by a transducive element, such as a microphone, is converted to a set of text words by a computer. These recognized words may then be used in a variety of computer software applications for purposes such as document preparation, data entry, and command and control. Improvements to speech dictation systems provide an important way to enhance user productivity. One style of improvement is to offer users the ability to make changes directly to dictated text, bypassing interaction with correction dialogs. Unless the system monitors changes and decides which are corrections to be sent to the speech engine for processing as corrections, and which are edits to be ignored by the system, the user will not receive the benefit of continual improvement in recognition accuracy that occurs when the engine receives correction information.
SUMMARY OF THE INVENTION
In a speech recognition system, a method of updating a language model for use when correcting dictated text comprises the steps of dictating a dictated word, providing a replacement word, and automatically comparing the dictated word to the replacement word using any suitable comparison means, such as using an algorithm to compare phonetics, grammar, spelling, or the context of surrounding words. If the comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word is a correction of a mis-recognition error rather than an edit, the method further comprises the step of determining if the replacement word is on an alternative word list. The alternative word can be preexisting or can be generated by any suitable method, including by the use of an algorithm which identifies words which have similar phonetics, grammar, and/or spelling. The method further comprises updating the language model without user interaction if the replacement word is on the alternative word list. If the replacement word is not on the alternative word list, dictated word digital information is compared to replacement word digital information, and the language model is updated if the digital comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word represents correction of a mis-recognition error rather than an edit.
The method can further comprise the steps of, prior to the digital comparison step, converting the audio of the dictated word into dictated word digital information and the text of the replacement word into replacement word digital information, and using the dictated word digital information and the replacement word digital information in the digital comparison step.
In the method, the replacement word can be generated by any suitable method, such as typing over the dictated word, pasting over the dictated word, or deleting the dictated word and replacing it with the replacement word. The dictated word can consist of a single word or a plurality words, but is generally a single word. Similarly, the replacement word can consist of a single word or a plurality of words, but is generally a single word.
According to a second aspect of the invention, the invention comprises a system for updating a language model during a correction session, which comprises a means for automatically comparing a dictated word to a replacement word using an suitable comparison means, such as using an algorithm to compare phonetics, grammar, spelling, and/or the context of surrounding words. If the comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word represents correction of a misrecognition error rather than an edit, the system further comprises a means for updating the language model without user interaction if the replacement word is on the alternative word list. The alternative word can be preexisting or can be generated by any suitable means, including by the use of an algorithm which identifies words which have similar phonetics, grammar, and/or spelling. If the replacement word is not on the alternative word list, the system further comprises a means for comparing dictated word digital information to replacement word digital information, and if the digital comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word represents correction of a misrecognition error rather than an edit, a means for updating the language model.
According to a third aspect, the invention comprises a machine readable storage, having stored thereon a computer program having a plurality of code sections executable by a machine for causing the machine to perform a series of steps. The machine readable storage causes the machine to perform the step of automatically comparing a dictated word to a replacement word using any suitable comparison means, including using an algorithm to compare phonetics, grammar, spelling, and/or the context of surrounding words. Further, the machine readable storage causes the machine to perform the steps of determining if the replacement word is on an alternative word list if the comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word represents correction of a misrecognition error rather than an edit, and updating the language model without user interaction if the replacement word is on the alternative word list. If the replacement word is not on the alternative word list, the machine readable storage causes the machine to perform the step of comparing dictated word digital information to replacement word digital information, and if the digital comparison is close enough, within a predetermined statistical quantity, to indicate that the replacement word represents correction of a misrecognition error rather than an edit, updating the language model.
REFERENCES:
patent: 5027406 (1991-06-01), Roberts et al.
patent: 5884258 (1999-03-01), Rozak et al.
patent: 5909667 (1999-06-01), Leontiades et al.
Nassiff Amado
Ortega Kerry A.
Dorvil Richemond
International Business Machines - Corporation
Nolan Daniel A.
LandOfFree
Smart correction of dictated speech does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Smart correction of dictated speech, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Smart correction of dictated speech will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2890383