INSTALLING
SIROCCO

Introduction

Sirocco is a project aiming at making available the source code of a large vocabulary speech recognition system based on continuous density Hidden Markov Models. The current package is the result of this effort to distribute such a code. The distribution contained in this directory contains the source codes (and the documentation, somewhat incomplete) of an LVCSR system (which means NO RESSOURCES AT ALL - such as acoustic models, lexicons, ...- are included in the distribution). The dat repertory solely contains toy ressources used by 'gmake test'.

The current distribution solely contains the decoder, ressources compilers, and tools to manipulate data files and word graphs. Here is a list of runtimes that will be built:

Downloads are Unix gzip archive files which contains the source files and the documentation as well as some (basic) installation instructions. Some more details can be found here under.

Compilation and Installation

Once the sirocco archive is uncompressed, just move to the Sirocco root directory and follow the following stages.

Content

The top level directory of the Sirocco contains the configure scripts and related files, README, COPYING... and the following subdirectories: Note that the content of auxdir/ and lib/ is distributed with Sirocco but *is not* part of Sirocco.

Compiling Sirocco

Compiling the package is done according to the standard (Gnu) procedure:

>> CFLAGS="-Wall -O3" ./configure
>> gmake

The above is the syntax for zsh and you may have to adapt for different shells. It is recommended that you compile with -O3 flag.

Compilation has been tested on Linux and on SunOS/Solaris systems. Note that you must have automake, Gnu CC and Gnu make installed on your system.

Installing Sirocco

>> gmake install

should install the binaries in the appropriate directory [usr/local/bin] is the default.

[home]

Last modified: Thu May 30 10:38:40 MET DST 2002 , by Guig.
(C) Sirocco consortium