Browse Prior Art Database

Bridging Speech Recognition and Natural Language Processing Subsystems

IP.com Disclosure Number: IPCOM000117231D
Original Publication Date: 1996-Jan-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 4 page(s) / 100K

Publishing Venue

IBM

Related People

Loring, K: AUTHOR [+2]

Abstract

Disclosed is a method for bridging the development of a speech recognition system based on Backus-Naur Form (BNF) grammar, such as the IBM Continuous Speech Series (ICSS) with the development of a parser generator also based on BNF grammar, such as the Yacc parser generator.

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

Bridging Speech Recognition and Natural Language Processing Subsystems

      Disclosed is a method for bridging the development of a speech
recognition system based on Backus-Naur Form (BNF) grammar, such as
the IBM Continuous Speech Series (ICSS) with the development of a
parser generator also based on BNF grammar, such as the Yacc parser
generator.

      The Figure is a block diagram of this method.  As shown in
section 1 of the figure, a speech recognition system 2 typically
accepts human speech in the form of an analog input from a microphone
3, converting the spoken message to digital information for further
processing into the from of a text string 4.  Phrases are accepted or
rejected based on a grammar 5 provided as input to the speech
recognition system 2.  While the text string 4 is useful as dictated
information, speech recognition systems of this kind do not
understand what has been said, and do not act on the statements.

      On the other hand, as shown in section 6, parsing is a process
of understanding language, in which discrete lexical units in an
input stream are logically grouped into statements within some
language.  A parser generator 7 is a utility, such as Yacc, which
generates code for a parser 8 when the developer supplies a language
specification, called a grammar 9, as well as functions that perform
actions based on specific phrases.  When the parsing code is compiled
and linked with a lexical scanner into an executable parser, the
parser 8 can accept a text string, such as the text string 4 from
speech recognition system 2, as input, performing actions based on
the text.

      The BNF grammars consist of productions, which are statements
that break down a phrase from a general form to a more specific form.
A phrase that cannot be broken down any further is a terminal;
otherwise it is a nonterminal.  A production in a BNF grammar is of
the following form:
             a -> {b|B}+
       where nonterminals are in lowercase,
         terminals are in uppercase,
         "a ->" means nonterminal "a" breaks down into the right side
              of the production,
         "{b|B}" means either nonterminal "b" or terminal "B", and
         "+" means one or more instances.

  ...