Browse Prior Art Database

Improved Correction of Stenographers' Errors through Audio Playback

IP.com Disclosure Number: IPCOM000104890D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 4 page(s) / 135K

Publishing Venue

IBM

Related People

De Gennaro, S: AUTHOR [+4]

Abstract

Stenographers typically introduce errors in the transcription of spoken text, for a variety of reasons. This disclosure presents a method to aid the stenographer in identifying and correcting his/her errors.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 41% of the total text.

Improved Correction of Stenographers' Errors through Audio Playback

      Stenographers typically introduce errors in the transcription
of spoken text, for a variety of reasons.  This disclosure presents a
method  to  aid  the  stenographer  in identifying and correcting
his/her errors.

      Stenographers  can  transcribe  spoken  text  with high speed
(200 words per minute).  This transcription consists of two steps:
first, stenographers type in  phonetic  utterance using  a  special
steno-machine  that  is connected  to a computer; secondly, this
steno  utterance is translated  into words.  This  can  be  done  in
two  ways:

1.  Manually: a stenographer reads encoded steno input  and
    translates  it into   words   (by   the  special  memorized
    table).

2.  Automatically: processed by a special program into  a computer
    and the best matched words are found automatically and displayed
    onto a screen and saved in a file.

      During such transcription, typing  errors  are  often
introduced by  stenographers themselves or by programs  that
translate the stenographer  phonetic  utterances  into  words.  Since
stenographers  are  focused  on  the  process  of  the transcription
during dictation or interpretation, they cannot monitor  errors
instantly  during  the  session.  They  can correct  errors  either
in  breaks  or  after  the  session is finished.  These  errors  can
be identified, in some cases, either using context, semantic and
grammar  errors,  or  as unrecognized  by  a machine phonetic
utterances.  In the process of correcting errors, stenographers
either rely on  their memory and context, or displaying a phonetic
utterance for a considered  error  output.  Both these correcting
methods are rather involved, cause significant delay in  cleaning  up
a transcription, and  fail  to  clean  some portion of errors.  Also
reviewing  the  transcript  at  a  later   time,   the stenographer
might  introduce  additional errors because of faulty use of context
or ambiguous phonetic strings.  Without the original audio for
reconstruction of the message, it  is not possible to detect such
errors.

      Conceptually, the method suggested here for dealing with steno
errors  involves the recording of the transcribed speech, indexing
the recorded speech by the transcribed text and  by playing  back
parts of a captured audio that aligned against found errors.

Below is a description  of  facilities  that  the  suggested method
requires.

1.  The transcribed text, which is displayed to the user.  This text
    might contain some words that were  not  processed by the
    automatic translation  system  (that  converts  steno utterance
    into words from the vocabulary).  These words  can be
    represented  on  a  display either by phonetic utterances that
    were produc...