![]() |
Sprex![]() |
||||||
|
| |||||||
|
News: Downloadable PDA Demos
Introduction |
GXC(1) gxc 1.2: Man Page Version 4 GXC(1)
NAME
gxc - grammar action compiler
SYNOPSIS
gxc -i file.gx [ -H file.h ] -l file.lat -c file.c [
-v ] [ -d ] [ -h ]
DESCRIPTION
gxc is a grammar action compiler. gxc converts the speci-
fied file.gx (grammar with actions) file into a recogni-
tion lattice file (HTK SLF format), named file.lat, as
well as a results-processing module, named file.c which
will carry out the associated actions when a recognition
result from the grammar is sent to it. file.c will
#include a supplied header file with declarations of any
functions called by actions in the .gx file.
For detailed documentation on building speech based appli-
cations using gxc, see the files gxc.STEPBYSTEP and
gxc.TUTORIAL, normally installed in the
/usr/share/doc/ansr-*/ directory.
OPTIONS
-i file.gx The named file is a grammar-with-actions,
or "gx" file. This is how you specify the possible word
sequences that the recognizer may expect, along with asso-
ciated actions that the results-processing agent should
carry out when handling any given recognition result.
-H file.h This additional input file is a C header
file which should be created while writing file.gx. You
can look at some sample header files that are provided for
the sample gx files in /usr/share/sprex/ansr/resources.
The filenames end in "_funcs.h". Note that the data types
of function arguments and function return values for those
functions match the declarations in the corresponding gx
files. file_funcs.h should contain appropriately typed
declarations for all of the functions called in the vari-
ous actions in file.gx, so that the results processing
agent can compile and call those functions correctly.
-l file.lat Into this output file will be written a
compiled grammar network representing all the word
sequences that could be spoken, as specified in file.gx.
Make this file available to sprecd so that it can recog-
nize the incoming audio stream against the grammar. Also,
check that the dictionary used by sprecd includes pronun-
ciations for each word in file.lat; lat2dict should help
with this.
-c file.c This output file is a C module that carries
out the actions associated with any given recognition
result, as specified in file.gx. It can be compiled
either as a standalone command-line program for testing
purposes which reads a file of text representing a recog-
nition result, or as an object file which can be linked to
sprecc or another client, enabling the client to call the
handle_results(fd) function, which reads results from the
(normally socket) file descriptor and carries out the
specified actions. In future versions of ANSR, this C
module may operate in the form of a separate daemon
program rather than as part of the client.
-v Verbose output
-d Debug information from .gx parsing process.
-h print usage information
EXAMPLES
To get a list of options and their explanation, use no
options:
example% gxc
Generate .lat and .c from .gx and .h
example% gxc -i file.gx -l file.lat -H file.h -c
file.c
Generate a dictionary for use with Entropic's HSGen pro-
gram, then generate 10 sample sentences from the lattice.
example% lat2dict -l file.lat -d file.dct
example% HSGen -n 10 file.lat file.dct | sed 's/[A-
Z0-9_$$.>]//g'
Compile agent into sprecc, run with sprecd to do the pro-
cessing.
example% cc -c file.c; cc -o sprecc sprecc file.o;
sprecd & sprecc
BUGS
None known. Please report bugs to tv@sprex.com. Your
feedback on this document is solicited.
SEE ALSO
lat2dict. Several additional docs are normally to be
found in /usr/share/doc/ansr-*, namely: gxc.TUTORIAL,
gxc.STEPBYSTEP, gxc.FAQERRORS, gxc.DESIGN, and of course
the sources. Start with gxc.STEPBYSTEP to get a quick
overview, then read gxc.TUTORIAL.
AUTHOR
Tom Veatch, Sprex, Inc. All rights reserved. gxc is
licensed software, to be used only under a license agree-
ment with Sprex.
Product of Sprex, Inc. 28 Aug. 2003 GXC(1)
|
||||||
Copyright © 1996-2005
Sprex, Inc.
All rights reserved. Sprex, Speech in the Network, TallyGram and ANSR are trademarks of Sprex, Inc.
|