uima-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marshall Schor <...@schor.com>
Subject Re: using multiple PEAR components in dynamically built pipelines
Date Thu, 17 Jan 2008 01:28:07 GMT
Hi Girish -

As I recall, you can only use a pear descriptor as a delegate of a 
containing aggregate.  The person who knows the most about this is away 
this week, so probably the definitive answer will come next week :-)


Chavan, Girish wrote:
> Hi All,
> I am trying to build a system where users will be able to build
> pipelines using any of the installed PEAR components and then save them
> for later use. From searching the mailing lists it has become apparent
> that you can use a PEAR component as an analysis engine. 
> Now, Looking at the SimpleRunCPE.java, and SimpleRunCPM.java code I can
> come up with two strategies to do this...
> Strategy 1: Create a CPE descriptor file based on the user selections of
> the PEAR components to be used and then build a CPE using the CPE
> descriptor similar to how it is done in SimpleRunCPE. 
> My question here is, can you specify a PEAR component descriptor file as
> a CasProcessor in the CPE descriptor file?
> Strategy 2: Create an Aggregate Analysis Engine descriptor that has the
> user selected PEAR components. Then specify that as the Analysis Engine,
> similar to how it is done in the SimpleRunCPM..java code.
> Again, can I use the PEAR descriptor file in the same fashion as I would
> use a TAE descriptor file in an Aggregate AE descriptor?
> Another alternative is to build the CPE programmatically, and add the
> PEAR components as CasProcessors programmatically. Basically do what the
> SImpleRunCPE might be doing behind the curtain
> (UIMAFramework.produceCollectionProcessingEngine() is a good place to
> look I presume)
> However the first two strategies are preferable as I would like the
> ability for users to build these pipelines in one session and then load
> and run them later in another session. For that I need to persist their
> selections in some form, and Id rather use a standard CPE descriptor or
> Aggregate AE descriptor file rather than introduce my own xml format.
> I hope I could put my point across clearly. Is anyone doing anything
> similar? Has anyone used multiple PEAR components in the same pipeline? 
> Any help would be much appreciated.
> Kind Regards,
> Girish Chavan
> Dept. of Biomedical Informatics,
> University of Pittsburgh

View raw message