Browse Prior Art Database

Distinction at Exit Between Text Editing and Speech Recognition Adaption

IP.com Disclosure Number: IPCOM000113908D
Original Publication Date: 1994-Oct-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 48K

Publishing Venue

IBM

Related People

Goldwyn, RM: AUTHOR

Abstract

Disclosed is a feature for a dictation system having text entry by speech and by keyboard, through which words entered by keyboard are used to generate new language model statistics and to modify the speech recognition process of the system, allowing the system to adapt to the speech of an individual user. This feature is employed when the user chooses to exit dictation, transferring text to another application, or when the user chooses to cut and paste text from the dictation system to another application.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 66% of the total text.

Distinction at Exit Between Text Editing and Speech Recognition Adaption

      Disclosed is a feature for a dictation system having text entry
by speech and by keyboard, through which words entered by keyboard
are used to generate new language model statistics and to modify the
speech recognition process of the system, allowing the system to
adapt to the speech of an individual user.  This feature is employed
when the user chooses to exit dictation, transferring text to another
application, or when the user chooses to cut and paste text from the
dictation system to another application.

      Text generated using the dictation system typically includes
tagged words, which have been supplied by the speech recognition
feature of the system, and untagged words, which have been supplied
by the user through the keyboard.  These untagged words may have been
supplied to correct words misinterpreted by the speech recognition
feature, or to provide other types of changes to the text.  At the
exit or cut and paste time described above, the untagged words are
used to generate new language model statistics, and the user is
offered an opportunity to pronounce each untagged word in turn.
Since the pronunciation of a known untagged word is used to adapt the
operation of the speech recognition feature to the user, the user is
expected to pronounce only those words which were typed due to a
failure of this feature to correctly recognize spoken material.

      Since this proc...