Automating cTAKES pipeline

[Durand Van Arnem <du...@hotmail.com>]

Hi,

I am getting started with cTAKES and need a nudge in the right direction.  I am developing a proof of concept, and what I'd like to accomplish is to have the cTAKES pipeline (like the AggregatePlaintextProcessor) running as a service so I can write code that will discover plain text documents, send them to the service, and get annotations back.  I just need to use cTAKES; I am not planning on extending it at this point.

I followed the cTAKES user installation guide and worked through the demos; they are straightforward and I can see all the annotations in the UI.

To automate the annotation process, it seems like I have these options:

1. Use the UIMA simple server to host cTAKES as in the example in the simple server user guide (http://uima.apache.org/downloads/sandbox/simpleServerUserGuide/simpleServerUserGuide.html).  I think this is my preferred solution; however, I am not sure how to generate a cTAKES PEAR.

2. Create a web service that calls cTAKES.jar (I am not sure what this would look like - do I create an app that loads the cTAKES AE XML file like this: http://uima.apache.org/d/uimaj-2.4.0/tutorials_and_users_guides.html#ugr.tug.application.instantiating_an_ae)?

   2a. Create a command-line interface that calls cTAKES.jar

3. Use cTAKES GUI (https://community.i2b2.org/wiki/display/NLPCTAKES/Home).  This appears to have done some of the work to wrap cTAKES; however, I can't tell if it's an active project (last update was at least 5 months ago, and it is using cTAKES 2.0).  Also, the instructions don't seem to match the download (I can't find a WAR in the download package).

Can you recommend one of these approaches or an alternative and point me in the right direction?

Thanks,
Durand 		 	   		  

RE: Automating cTAKES pipeline

["Chen, Pei" <Pe...@childrens.harvard.edu>]

Hi Durand,
Thank you for your interest.  Another option may be to use UIMA-AS (UIMA Async Scaleout).
Even though it was designed to as a scale out solution, it has all of the necessary components to wrap cTAKES into a service architecture.

We will be exploring this and hopefully add a cTAKES-service as a scaled out option in the sandbox very soon.  Is this something you would be also interested in contributing?
Essentially, cTAKES-service would just take in an expected xml (xmi cas?) and return an xml (xmi cas?).  Applications would just need to write code to feed in the documents and consume the xml.

Re: 3) yes, the ctakes-gui is currently in Apache cTAKES' sandbox now.  We will most likely add it as an option component and hopefully reconfigure it to take advantage of the above architecture in the future.

> -----Original Message-----
> From: Durand Van Arnem [mailto:durand_v@hotmail.com]
> Sent: Thursday, November 29, 2012 1:17 PM
> To: ctakes-user@incubator.apache.org
> Subject: Automating cTAKES pipeline
> 
> 
> Hi,
> 
> I am getting started with cTAKES and need a nudge in the right direction.  I
> am developing a proof of concept, and what I'd like to accomplish is to have
> the cTAKES pipeline (like the AggregatePlaintextProcessor) running as a
> service so I can write code that will discover plain text documents, send them
> to the service, and get annotations back.  I just need to use cTAKES; I am not
> planning on extending it at this point.
> 
> I followed the cTAKES user installation guide and worked through the demos;
> they are straightforward and I can see all the annotations in the UI.
> 
> To automate the annotation process, it seems like I have these options:
> 
> 1. Use the UIMA simple server to host cTAKES as in the example in the simple
> server user guide
> (http://uima.apache.org/downloads/sandbox/simpleServerUserGuide/simpl
> eServerUserGuide.html).  I think this is my preferred solution; however, I am
> not sure how to generate a cTAKES PEAR.
> 
> 2. Create a web service that calls cTAKES.jar (I am not sure what this would
> look like - do I create an app that loads the cTAKES AE XML file like
> this: http://uima.apache.org/d/uimaj-
> 2.4.0/tutorials_and_users_guides.html#ugr.tug.application.instantiating_an_
> ae)?
> 
>    2a. Create a command-line interface that calls cTAKES.jar
> 
> 3. Use cTAKES GUI
> (https://community.i2b2.org/wiki/display/NLPCTAKES/Home).  This appears
> to have done some of the work to wrap cTAKES; however, I can't tell if it's an
> active project (last update was at least 5 months ago, and it is using cTAKES
> 2.0).  Also, the instructions don't seem to match the download (I can't find a
> WAR in the download package).
> 
> Can you recommend one of these approaches or an alternative and point me
> in the right direction?
> 
> Thanks,
> Durand