Browse Prior Art Database

Introducing Grammar Checking into an Automatic Dictation System

IP.com Disclosure Number: IPCOM000109563D
Original Publication Date: 1992-Sep-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 2 page(s) / 86K

Publishing Venue

IBM

Related People

Chanod, JP: AUTHOR [+3]

Abstract

Herein disclosed is an automatic dictation system (ADS) enhanced by coupling a grammar checker as a post-processor. ADS are nowadays powerful and reliable. They can handle large vocabularies (20,000 forms), but they still produce errors due to some inadequacies of the underlying linguistic models. The grammar checker is able to diagnose and correct grammatical errors produced by the initial output of the ADS, and thus it improves its overall performance.

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

Introducing Grammar Checking into an Automatic Dictation System

       Herein disclosed is an automatic dictation system (ADS)
enhanced by coupling a grammar checker as a post-processor.  ADS are
nowadays powerful and reliable. They can handle large vocabularies
(20,000 forms), but they still produce errors due to some
inadequacies of the underlying linguistic models. The grammar checker
is able to diagnose and correct grammatical errors produced by the
initial output of the ADS, and thus it improves its overall
performance.

      The ADS is implemented on a personal computer IBM PS/2* or IBM
RISC System/6000* to which are added a vocal I/O card as well as a
specialized card equipped with two microprocessors, which provide the
needed power for the decoding algorithms.

      The voice signal is submitted to a chain of signal processing,
in order to extract acoustic parameters from the sound wave.

      Two passes of acoustic evaluation are performed: a relatively
gross pass selects a reduced list of candidate words; this list is
further reduced thanks to the language model so that only a small
number of remaining candidates are submitted to a second, more
precise, acoustic pass.

      The decoding algorithm, relying on probabilistic models,
determines the more likely uttered sequence of words.  It works from
left to right by combining the various scores estimated by the
acoustic and linguistic models, according to a so-called stack
decoding strateg...