SPADE, Release Developer 0.9  
	(c) 2003 University of Southern California
 		written by Radu Soricut


SPADE (Sentence-level PArsing for DiscoursE) is a discourse parser at sentence 
level written by Radu Soricut at USC/ISI. You can find details about the 
approach implemented by SPADE in the paper: 

Radu Soricut and Daniel Marcu (2003). 
Sentence Level Discourse Parsing using Syntactic and Lexical Information. 
Proceedings of the Human Language Technology and North American Association 
for Computational Linguistics Conference (HLT/NAACL), May 27-June 1, 
Edmonton, Canada. 

SPADE assumes as input a one-sentence-per-line file, and the output is one 
discourse parse tree per sentence. The algorithm uses syntactic parse trees, 
and needs to call Charniak's syntactic parser. See the bottom of this page 
for instructions on how to integrate Charniak's parser with SPADE. 
Both Charniak's parser and SPADE use SUN executables; SPADE also needs 
Perl 5.0 (or newer). 

An embbeded feature of SPADE is the discourse segmenter. 
You can run SPADE with the -seg-only option; in this case, the output is 
one segment (elementary discourse unit) per line, with end-of-sentence 
boundaries inserted at the original (input) sentence boundaries. 

To run SPADE, you need to download Charniak's syntactic parser, available at
http://www.cs.brown.edu/people/ec/#software
After you install Charniak's parser on your machine, you must edit the file 
spade.pl (in the bin/ directory); in line 3, set the value of $CHP to the 
directory path for Charniak's parser.

IF you need to create executables for an architecure different than
SUN, you should do the following (otherwise, skip these steps):

% cd bin/
% rm dependencies edubreak
% make

After the new executables are created, you are ready to run SPADE.

Run SPADE:

% spade.pl [-seg-only] one-sentence-per-line-file

Example:

% spade.pl ../example/wsj_0616.sent