Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1998-11-03
2001-04-17
Hudspeth, David (Department: 2741)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S231000, C704S270000, C704S251000, C379S100080, C379S093240, C455S412100
Reexamination Certificate
active
06219638
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to editing systems for voice recognition and, more particularly, to a system and method for editing messages transcribed from speech from a telephone.
2. Description of the Related Art
Advances in personal communications in recent years have led to information being transmitted through a variety of channels to users, for instance speech, multi-media (figures and speech), text (e-mail, pagers), etc. Due to these advances, there has arisen the concept of unified whereby the messages received by a user through various media are stored in a single repository and can be retrieved or searched by the user at his/her convenience. Further, it may be the case that the user has only a personal digital assistant (PDA) with very limited capabilities through which to retrieve his messages. In general however, even the simplest of PDA's will support the reception of text, though it may not support the reception of multimedia signals. Consequently, it may be necessary to convert speech and multimedia signals into text so that the signals can be easily accessed. This also has implications on the bandwidth requirements for communication—text signals require less bandwidth than speech for transmission.
Voicemail is a commonly used messaging system wherein the speech of a person is recorded and subsequently played back by the recipient of the message. Hence, an important component of unified messaging is the capability to convert such messages into text. This can of course be done by using automatic speech recognition algorithms. However, voicemail messages typically represent spontaneous speech recorded over an unknown (the caller who is leaving the message may be halfway around the earth or next door) telephone bandwidth channel, and hence represent a very challenging task for automatic speech recognition systems. there is the danger of the transcribed text being so full of errors that the recipient of the message may not be able to decipher the message at all. Hence, it is advantageous to incorporate some form of feedback mechanism whereby the person leaving the message can check the quality of the transcription and correct it if necessary.
Therefore, a need exists for an interactive system and method for converting speech data into text and incorporating the feature of correction of the transcribed text by voice.
SUMMARY OF THE INVENTION
A messaging system, in accordance with the present invention, for receiving speech and converting the speech to text includes a first server for receiving speech input by a user, a speech recognition system for converting the speech to text, a speech synthesizer for converting the text to speech for playing back the synthesized speech for correction by the user and a correction mechanism for enabling the user to correct the speech such that the corrected speech is provided as text for transmittal over a communication system. other embodiments, the text for transmittal over the communication system may include transmittal to one of a pager, email and fax. The correction mechanism may prompt the user to select portions of the speech input for correction. The speech recognition server may provide diagnostic data to the correction mechanism to indicate portions of the speech input to be corrected. The correction mechanism may prompt the user to rerecord portions of the speech input for correction. The system may further include a language translation server for converting the speech input to text for transmittal over the communication system in a different language. The system may further include a speaker identification server for identifying the user and for adjusting speech recognition models for speech recognition by the speech recognition server.
A method for correcting messages for a universal messaging system includes the steps of recording an audio message, transcribing the message to text using a speech recognition system, providing speech in accordance with the transcribed text for playing back the message for correction, identifying portions of the message to be corrected, correcting the message by re-recording the portions and outputting the text over a communication system.
A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for providing corrections to messages in a universal message system, includes the method steps of recording an audio message, transcribing the message to text using a speech recognition system, providing speech in accordance with the transcribed text for playing back the message for correction, identifying portions of the message to be corrected, correcting the message by re-recording the identified portions and outputting the text over a communication system.
In alternate methods which may be executable by the program storage device, the audio message is preferably recorded by telephone. The step of identifying portions of the message to be corrected may include the step of providing diagnostic data from the speech recognition server for determining a likelihood of correctness of the portions of the message. The step of identifying portions of the message to be corrected may include the step of listening to the played back message and selecting portions to be corrected. The step of correcting the message by re-recording the identified portions may include the steps of re-recording portions of the message, converting the re-recorded portions to revise the text using the speech recognition server, playing back speech of the re-recorded portions in accordance with the revised text and if acceptable, approving the portions of the message. The step of recording the message in a one of a plurality of languages may be included. The step of outputting the text in a one of a plurality of languages over the communications system may be included. The steps of identifying a user associated with speech recognition models and applying the models to recognize the audio input of the user may be included.
REFERENCES:
patent: 5051924 (1991-09-01), Bergeron et al.
patent: 5875448 (1999-02-01), Boys et al.
patent: 5920835 (1999-07-01), Huzenlaub et al.
VoiceAssist™ (Creative Labs, “User's Guide,” Jul. 1993.*
Talk> To Plus™ (Dragon Systems, “User's Guide”, ©1992-1993).*
Rose et al (Richard C. Rose, Douglas A. Reynolds, “Text independent speaker identification using automatic acoustic segmentation,” International Conference on Acoustics, Speech, and Signal Processing, Apr. 1990).
Nahamoo David
Padmanabhan Mukund
Picheny Michael
Roukos Salim
F. Chau & Associates LLP
Hudspeth David
International Business Machines - Corporation
Nolan Daniel A.
LandOfFree
Telephone messaging and editing system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Telephone messaging and editing system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Telephone messaging and editing system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2502998