Browse Prior Art Database

Method for Logging Edits and Deferring Corrections in a Speech Recognition Dictation System

IP.com Disclosure Number: IPCOM000123222D
Original Publication Date: 1998-Jul-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 2 page(s) / 85K

Publishing Venue

IBM

Related People

Ballard, B: AUTHOR [+3]

Abstract

Disclosed is a method allowing the user of a speech recognition system to delay various activities associated with the correction of speech recognition errors until a convenient time after the completion of a dictated document.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Method for Logging Edits and Deferring Corrections in a Speech Recognition
Dictation System

   Disclosed is a method allowing the user of a speech
recognition system to delay various activities associated with the
correction of speech recognition errors until a convenient time after
the completion of a dictated document.

   In a conventional speech recognition system, the correction
of recognition errors is used to improve the accuracy of
recognition.  It is suggested that users should dictate a paragraph
and then correct the errors.  However, since the correction of
recognition errors is more time consuming than making changes by
editing the text, many users fix the document by editing, so that the
speech recognition improvements with recognition correction are not
achieved.

   The present method provides an option, which can be set
by the user, causing the system to save all dictated text
corresponding to text replaced by the user during an editing
process.  With this option, the user edits text normally, with the
system saving the old text, together with its corresponding dictated
audio, including a number of words before and after the correction to
provide context, in a correction file.  When the user initially
selects this option, the system saves the file name and a short
description provided by the user in the correction file.

   FIG. 1 is an example of screen provided by the recognition
improvement tool in a first version of this method.  At any time,
the user can open this tool, specifying the correction file desired.
While the screen initially highlights a word that has been changed by
editing, the user can select any number of words to correct.  If the
correct word is not in the alternatives listed on the screen, the
user types the correct word.  After he has corrected the word, he can
re- dictate the text to see if the system recognizes the word
correctly.  If a new recording is required, the system prompts the
user to record the word.  This screen includes a number of
cursor-selectable buttons.  After the user selects one of the
alternative words by typing a number, selecting the "Correct" button
causes the change to be made, with the next word to be corrected then
being displayed.  Selecting the "Skip" button causes the system to
skip the present correction without making a change, proceeding to
the next correction.  Selecting the "Previous" button causes the
system to retur...