Corpora: Alembic Workbench Announcement

David S. Day (day@linus.mitre.org)
Fri, 24 Jul 1998 14:45:57 -0400 (EDT)

I am sending you this message because you have either downloaded a
copy of the Alembic Workbench from the MITRE Corporation's web site
(www.mitre.org, www.mitre.org/technology/alembic-workbench), or
otherwise indicated an interest in this software. I would like to
make you aware of three new developments.

---------------- * ----------------

The first bit of news is that we have established a new mailing list:

awb-users@linus.mitre.org

This list is for use by anyone who wishes to discuss with other users,
or with MITRE developers of the Workbench, any issues regarding
release status, bugs, desired enhancements, useful tag-preference
files, etc., etc. The Alembic Workbench is being used by a number of
academic, commercial and governmental organizations, and we hope this
list will encourage users to exchange information that they have found
useful in developing textual corpora using this tool. The mailing
list is completely un-moderated. To subscribe, send email to

awb-users-request@linus.mitre.org

and have the body of the message contain the following:

subscribe awb-users

(The subject header of the message is ignored.)

+---------------------------------------------------------------------+
| PLEASE NOTE: No one has been added to the awb-users list without |
| their consent. If you wish to subscribe to this mailing list, you |
| must yourself send the subscribe message described above. The |
| message you are currently reading has been sent using information |
| provided by the download form at our web site, or via some other |
| communications you have had with someone in MITRE's NLP group. |
+---------------------------------------------------------------------+

---------------- * ----------------

The second bit of news is that we have built a new Windows95/NT
version of the Alembic Workbench corpus annotation tool and it is
available for downloading from our web site along with the other
versions available there (for Solaris, SunOS, and Linux). This
installs very easily (see the installation instructions for special
restrictions on folder names). This is a very new version, with a
very small user base, so we encourage any users to report bugs to us
either using the awb-users mailing list, or directly to one of the
contacts listed on our web pages.

There are also now two new abbreviated URLs for reaching our web site:

www.mitre.org/technology/nlp
-- An index into MITRE's natural language processing group

www.mitre.org/technology/alembic-workbench
-- An index into MITRE's Alembic Workbench-related pages

The old URLs are still correct, and are not scheduled to be changed.

---------------- * ----------------

Finally, we would like to point out that the latest full distributions
of the Alembic Workbench, Version 2.16 and greater, include an updated
version of our multi-lingual natural language processing system,
Alembic. This new version incorporates two important changes. One is
that we have adopted a parameter setting convention that allows almost
all of the system's capabilities and processing stages to be
controlled by the user via declarations that can be saved in "spec"
files. These spec files can inherit, via INCLUDE declarations,
settings from other, more general spec files. This enables
specialized spec files to be quite succinct, since they only need
specify those parameter settings that are different from the inherited
files. Another important enhancement is a new unix-level calling
facility, process-doc, that allows Alembic to be called from any unix
shell (running on a Sun Solaris or Sun OS machine). To see the
argument pattern for process-doc, use the -h as in:

unix> process-doc -h

Here's an example of how to use process-doc to perform "named entity"
(NE) tagging in the manner of the Seventh Message Understanding
Conference (MUC7):

unix> process-doc $ALEM/specs/muc7-specs.spec \
prelembic.in $AWB/data/muc6-ft-1-5.sgml \
prelembic.out $AWB/data/muc6-ft-1-5.tag \
lisp.in $AWB/data/muc6-ft-1-5.tag \
ne.out $AWB/data/muc6-ft-1-5.ne \
lisp.phases NE

The example above should run exactly as stated using any properly
installed Alembic/Alembic Workbench system. The NE-tagged data should
be found in the file $AWB/data/muc6-ft-1-5.ne. If you wish to know
more about using or customizing Alembic, please feel free to call or
write to Marc Vilain (mbv@mitre.org, 781-271-2151) or me.

Thank you very much.

- David Day

+-----------------------+---------------------------------------------+
| Dr. David S. Day | WWW: http://www.mitre.org/ |
| MS K329 | http://www.mitre.org/technology/nlp |
| The MITRE Corporation | Intelligent Information Access (G6H) |
| 202 Burlington Road | Artificial Intelligence Center, AISC (G60) |
| Bedford MA 01730 USA | Center for Integrated Intelligence Systems |
| Phone: (781) 271-2854 | Fax: (781) 271-2352 Email: day@mitre.org |
+-----------------------+---------------------------------------------+