Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-03-08
2001-06-26
Dorvil, Richemond (Department: 2741)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S255000
Reexamination Certificate
active
06253177
ABSTRACT:
CROSS REFERENCE TO RELATED APPLICATIONS
(Not Applicable)
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
(Not Applicable)
BACKGROUND OF THE INVENTION
1. Technical Field
This invention relates to the field of computer speech recognition and more particularly to a method and system in which a speech recognition application can automatically determine whether a user implemented amendment to dictated text should serve as a basis to update a language model.
2. Description of the Related Art
Speech recognition is the process by which an acoustic signal received by microphone is converted to a set of 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. Speech recognition is generally a difficult problem due to the wide variety pronunciations, individual accents and speech characteristics of individual speakers. Consequently, language models are often used to help reduce the search space of possible words and to resolve ambiguities as between similar sounding words. Such language models tend to be statistically based systems and can be provided in a variety of forms. The simplest language model can be specified as a finite state network, where the permissible words following each word are given explicitly. However, more sophisticated language models have also been developed which are specified in terms of a context specified grammar.
Since each particular user will tend to have their own style of speaking, it is important that the attributes of such speaking style be adapted to the language model. By continuously updating the language model, it is possible to improve the overall accuracy of the speech recognition process for that user and thereby permit greater efficiencies. Accordingly, it is desirable to update the language model for each particular user on a regular basis. This updating process may be accomplished by various means. For example, the speech recognition application may be provided with sample text in some instances as examples of a particular users' speaking style. This information may thereafter be used, if suitable, to update the language model. Another method for updating the language model involves making note of user“corrections” to misrecognized words in dictated text. However, this method can be cumbersome as it is necessary for the user to distinguish between amendments to text which are corrections to misrecognized words as compared to those which are merely additions or revisions to text. Corrections to misrecognized words are appropriately considered for updating the language model whereas other additions or revisions to text for other reasons are not necessarily so. The difficulty for the speech recognition application is distinguishing one type of amendment to the document from another.
SUMMARY OF THE INVENTION
The invention concerns a method and system for automatically determining whether to update a language model in a speech recognition system based upon user amendments to text in a dictated word processing document. The method involves monitoring one or more text-amending inputs to an application program for amending the text. Text-amending inputs are any kind of user initiated input which can be used to amend a word processing document. Examples of such text-amending inputs include keyboard entries to type in text, mouse movements and clicks which cause selection or de-selection of certain text, backspace and delete key operations which cause text to be removed from the word processing document and any other such similar operations which a user can perform in the course of amending a word processing document text.
The second step involves processing performed in response to the monitoring step to determine whether the text-amending input corrects a misrecognized word in the dictated text. This is accomplished by applying certain heuristic rules which permit such a determination. For example, the processing step can involve determining whether the text amending input has caused dictated text to be selected and determining whether new text has been inserted within the selected text.
Alternatively, or in addition thereto, the processing step can involve determining whether a text amending input has caused dictated text to be selected; determining whether the selected text has been deleted by the text amending input; determining whether the cursor has not been moved; and determining whether a text-amending input has caused new text to be inserted in place of the deleted text. Where these user inputs have all occurred, the system can reasonably conclude that a misrecognized word has been deleted and replaced with a correct word.
According to another aspect of the invention, the processing step can involve determining whether a text amending input has caused the positioning of a cursor over an existing word in the dictated text and determining whether the existing word has been overwritten as a result of a subsequent text amending input. Based upon such inputs, the system can reasonably conclude that a misrecognized word has been deleted and replaced with a correct word.
According to still another aspect of the invention, the processing step can involve determining whether the text amending input is the addition of new text to the dictated text and determining whether a further text amending input is the selection and deletion of text immediately contiguous to the new text. This would be an indication that a misrecognized word has been replaced with a correct word.
According to yet another aspect of the invention, the processing step can involve determining whether the text amending input is the addition of new text to the dictated text, and determining whether a further text amending input is the deletion of text immediately contiguous to the new text by operation of either a delete key or a backspace key.
REFERENCES:
patent: 5465317 (1995-11-01), Epstein
patent: 5680511 (1997-10-01), Baker et al.
patent: 5852801 (1998-12-01), Hon et al.
patent: 5970460 (1999-10-01), Bunce et al.
patent: 5995928 (1999-11-01), Nguyen et al.
patent: 6081772 (2000-06-01), Lewis
patent: 0841655 A2 (1998-05-01), None
patent: 0840288 A2 (1998-05-01), None
patent: 0840289 A2 (1998-05-01), None
patent: 2347774 (2000-09-01), None
IBM Technical Disclosure Bulletin. Improved Correction of Speech Recognition through Audio Playback. Jun. 1993.*
IBM Technical Disclosure Bulletin. Improved “Added-Word” User Interface Using Integrated Automatic Speech and Handwriting Recognition. Feb. 1993.
Ballard Barbara
Lewis James R.
Dorvil Richemond
International Business Machines Corp.
Senterfitt Akerman
LandOfFree
Method and system for automatically determining whether to... 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 system for automatically determining whether to..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for automatically determining whether to... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2520203