Mailing-List: contact user-help@uima.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@uima.apache.org
Received-SPF: pass (nike.apache.org: domain of kottmann@gmail.com designates
 74.125.82.175 as permitted sender)
Message-ID: <4EEB1F66.3020105@gmail.com>
Date: Fri, 16 Dec 2011 11:37:26 +0100
From: =?ISO-8859-1?Q?J=F6rn_Kottmann?= <kottmann@gmail.com>
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6;
 rv:8.0) Gecko/20111105 Thunderbird/8.0
MIME-Version: 1.0
To: user@uima.apache.org
Subject: Re: Merging CASes
References: <9D1B75EABCEEBF40A683AE4A1F9FDFBD046264DD@swmmbx00.bk.bka.bund.de>
In-Reply-To: 
 <9D1B75EABCEEBF40A683AE4A1F9FDFBD046264DD@swmmbx00.bk.bka.bund.de>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

At OpenNLP we take the gold data, extract the text, run
it through a one of our components and compute the
scores on the fly.

In your case I would do it similar. I would take the XMI files
generated from the gold standard corpus, copy the text
and other input data from the initial view to a second view.
Now you can run the tagger over the second view, after you
did this an additional analysis engine could compute your
performance scores.

J�rn

On 12/16/11 7:18 AM, Armin.Wegner@bka.bund.de wrote:
> Hi,
>
> I have got an XMI-file with the gold-standard as initial view and
> another XMI-file with tagger output as inital view. Now I would like to
> combine both into one CAS, such that the gold-standard is in the gold
> view and the tagger ouput in the tagger view. Than, I can easly compute
> the F-values. Any ideas how to do that? Is it possible to deserialze two
> CASes in such a way?
>
> The problem is that I have to use the two files. With one file and some
> annotators one could do it like this: Use an collection reader to read
> the gold-standard from the XMI-file. Create a new view for the tagger.
> Copy the sofa from the gold view to the new tagger view. Use an
> annotator to tag the tagger view sofa. You could repeat the last three
> steps for different taggers. Than you would have all annotations for all
> taggers in one huge CAS. What do you think about this?
>
> Greetings,
>
> Armin