Newsgroups: comp.ai.nat-lang
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!godot.cc.duq.edu!newsgate.duke.edu!zombie.ncsc.mil!nntp.coast.net!news.kei.com!newsfeed.internetmci.com!iol!tank.news.pipex.net!pipex!dish.news.pipex.net!pipex!canon.co.uk!news
From: David Elworthy <dahe@cre.canon.co.uk>
Subject: Re: HMM PoS-tagging
X-Nntp-Posting-Host: scutter
Content-Type: text/plain; charset=us-ascii
Message-ID: <31B82C81.13728473@cre.canon.co.uk>
Sender: news@canon.co.uk (USENET News)
Content-Transfer-Encoding: 7bit
Organization: Canon Research Centre Europe Ltd
References: <4p95iv$f35@highway.leidenuniv.nl>
Mime-Version: 1.0
Date: Fri, 7 Jun 1996 13:20:01 GMT
X-Mailer: Mozilla 2.01 (X11; I; SunOS 4.1.4 sun4m)
Lines: 16

Stephan Raaijmakers wrote:
> 
> Does anyone have practical experience in building a HMM PoS-tagger
> (preferably trained using Baum-Welch)? If so, would you like
> to share some experiences/thoughts on this?

Have a look at my paper "Does Baum-Welch Re-estimation Help Taggers?" in
the 1994 Applied NLP conference (Stuttgart). It contains some
experiments on BW and specific suggestions for how to use it
effectively. You can get it from the cmp-lg eprint server.

_______________________________________________________________________
David Elworthy <dahe@cre.canon.co.uk>
Canon Research Centre Europe Ltd., Guilford, Surrey, UK
Phone: +44 1483 448844; Fax: +44 1483 448845;
http://www.cre.canon.co.uk/
