SFST README
Note on this version
====================
This is the 1.2 version taken from the original site but with modified Makefile.
The modifications cause the buildprocess to automatically build and install the
libsfst.so library with headers (this was not the case in the original version).
The modifications were made quickly and may contain bugs. The modifications were
made by Adam Radziszewski to make it easier to install the SFST plugin for the
MACA system.
Installation (Unix and Linux only)
============
(Thanks to Robert Rathbone for suggestions which improved this README file.)
In order to install SFST you need the GNU version of the make program,
the GNU C++ compiler and the standard C++ library including the SGI
extensions hash_set and hash_map.
You might additionally need one or more of the following tools:
1) The "Flex" scanner generator which can be downloaded from:
http://flex.sourceforge.net. Flex relies on m4 to function properly.
2) The "m4" macro processor which must be installed along with Flex.
Download from: http://www.gnu.org/software/m4
3) The "Bison" parser generator available from http://www.gnu.org/software/bison
4) The "Ncurses" programming library available from http://www.gnu.org/software/ncurses
5) Once they are installed you should be able to install SFST without
any further problems.
After unpacking the software package, change to the source
directory and call
> make
to compile the tools. Then call
> make install
to install the tools in /usr/local/bin. Finally call
> make maninstall
to install the manpages in /usr/local/man/man1 and you are done.
The subdirectory data contains a simple example of an English
morphological analyser, the source code of the German SMOR morphology
(with just a few sample lexicon entries), and the general morphology
XMOR which may be used as a starting point for the development of a
computational morphology.
Problems with readline
----------------------
If you get an error message from the C++ compiler saying that it is
unable to find "readline/readline.h", you should install the GNU
readline package which is freely available at
http://directory.fsf.org/GNU/readline.html
Problems with hash_set and hash_map
-----------------------------------
If you get a message saying that hash_set and hash_map are not found,
you should uncomment the following line in the Makefile
#SGILIB = -DSGIext
If that does not help, make sure that hash_set and hash_map are
contained in your C++ libraries. Check in which subdirectory the
include files for hash_map and hash_set are located and to which name
space they are added. Then edit in the file alphabet.h the lines
starting with
#ifdef SGIext ...
appropriately.
Usage
=====
See the manual SFST-Manual.pdf in the doc subdirectory and the man
pages for more information on the SFST tools. doc/SFST-Tutorial.pdf
explains how computational morphologies are implemented in SFST.
If you want to implement your own tools based on the SFST code, have a
look at fst-infl.C and fst-infl2.C. They show how analysers are
implemented with the standard (fst-infl) and the compact (fst-infl2)
transducer format.
Author
======
The SFST tools have been implemented by Helmut Schmid, Institute for
Computational Linguistics, University of Stuttgart, Germany and they
are available under the GNU public license version 2 or higher.
Bug reports
===========
Please send bug reports and other feedback to schmid@ims.uni-stuttgart.de.
