CWB beta version (pre-3.0), binaries for various platforms
----------------------------------------------------------

To install the beta version, copy the appropriate archive for
your operating system (Linux, Solaris, IRIX, AIX) into 
/usr/local (recommended directory) or any other directory 
(you'll then have to adjust the path), and un-gzip and
un-tar it.

Available versions:

cwb-2.2.b72-sparc-solaris.tar.gz	(SUN Solaris 2.6+)
cwb-2.2.b72-i386-linux.tar.gz		(Linux/i386 2.0+)

"Unsupported" versions:

cwb-2.2.b17-mips-irix.tar.gz		(SGI IRIX)
cwb-2.2.b17-powerpc-aix.tar.gz		(IBM AIX)


For example, if you're working on a Linux machine, 
type the following:

    gunzip cwb-2.2.b72-i386-linux.tar.gz
    tar xvf cwb-2.2.b72-i386-linux.tar

    (if you want to see the contents of the tar file first use "tvf")

This creates, amongst others, a directory bin/ which contains the binaries
for CQP, the Corpus Query Processor, and encode/makeall, which are used for 
encoding corpora.

To find a corpus, CQP uses an environment variable $CORPUS_REGISTRY. This
has to point to a directory registry/ where the corpora on your system are 
defined. 

You'll find an example corpus in this directory, the file "susanne.tar.gz", 
and the registry file for it, "susi".


Here's an example on how to proceed to query this corpus:

    gunzip susanne.tar.gz
    tar xvf susanne.tar

This will create a subdirectory susi/ in the current directory. Now 
create another subdirectory registry/ . 

    mkdir registry

Copy the file "susi" to the subdirectory registry/ and change the
path after the string "HOME" to the current path followed by "susi".

The file will then look like this (without the "-----"):
-----
NAME "Susanne-Corpus"
ID susi
HOME <YOUR_PATH>/susi

ATTRIBUTE word
ATTRIBUTE tag
ATTRIBUTE pos
ATTRIBUTE lemma
-----

Set the environment variable $CORPUS_REGISTRY to the registry 
subdirectory.

    setenv CORPUS_REGISTRY "<YOUR_PATH>/registry"

    ( alternatively: CORPUS_REGISTRY=<YOUR_PATH>/registry )

(If that doesn't work you can call cqp with the option 
"-r <YOUR_PATH>/registry").

Now CQP will find the registry subdirectory, and because the path to the
actual corpus data is given in the file "susi" in the registry subdirectory,
the corpus can be queried. 


Here's an example for some simple cqp commands:

    bin/cqp -e
    [no corpus]> SUSI;
    SUSI> "interesting";                                
    18480:  the Legislature with an  <interesting> dilemma . Since the cons
    55388: llum renaissance . It is  <interesting> , however , that despite
    59525:  « Why , it 's extremely  <interesting> . But I - - would never 
    ...

    SUSI> set Last collocate leftmost [pos="V.*"] within 5 words right from match exclusive; 

    (For details on the syntax please have a look at the manual,
    http://www.ims.uni-stuttgart.de/projekte/CorpusWorkbench/UsersCorner.html,
    or at the file new_features.txt)

    SUSI> delete without collocate;
    SUSI> cat;
    SUSI> group Last target word;
    SUSI> cat Last > "/tmp/query_results";
    SUSI> exit

Calling cqp with the option "-h" will show you all the command line
switches. If you know the name of the corpus you're going to work with 
you can for example type "cqp -D SUSI" (the name of the corpus has to 
be typed in capital letters...). The "-e" switch allows for command line 
editing.

If there are any questions or something doesn't seem to work don't
hesitate to contact us,

Arne Fitschen    fitschen@ims.uni-stuttgart.de
Stefan Evert     evert@ims.uni-stuttgart.de

Stuttgart, 20.12.2001,
Arne Fitschen/Stefan Evert