Mailing-List: contact user-help@uima.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@uima.apache.org
MIME-Version: 1.0
In-Reply-To: <C84DD3AC-0D5C-4D9B-A0C2-D88559F42AFB@apache.org>
References: 
 <CAMjp+0fMcbbdogwDxF+f2oKu=aNu0oA+xnkK=FQxzebb7WKTmw@mail.gmail.com>
	<C84DD3AC-0D5C-4D9B-A0C2-D88559F42AFB@apache.org>
Date: Sat, 29 Aug 2015 05:59:05 -0400
Message-ID: 
 <CAMjp+0d9RLY+UifS7D-q9yB4nK6tAzL04iNS1R5Lu_2PvE3drQ@mail.gmail.com>
Subject: Re: Exposing a View to Downstream AEs
From: Matthew DeAngelis <ronin78@gmail.com>
To: user@uima.apache.org
Content-Type: multipart/alternative; boundary=089e01493fa82d0d64051e704085

--089e01493fa82d0d64051e704085
Content-Type: text/plain; charset=UTF-8

Hi Richard,

Thank you so much, this works like a charm! I knew it was about sofa
mappings, but I was having a lot of trouble figuring out how to do that
mapping in the Component Descriptor Editor. The uimaFIT approach is much
nicer, and your guidance on how to do the mappings is much appreciated.


Regards,
Matt

On Sat, Aug 29, 2015 at 4:27 AM, Richard Eckart de Castilho <rec@apache.org>
wrote:

> Hi Matt,
>
> what you are looking for is called "sofa mapping". On the level of an
> aggregate, you can define view names and map these names to local
> names in each of the AEs.
>
> E.g. if you have one AE that produces a view it calls "output" and
> another one that expects its data in "_InitialView", then you could
> set up a mapping like
>
>  AAE "data" -> AE1 "output"
>  AAE "data" -> AE2 "_InitialView" (CAS.NAME_DEFAULT_SOFA)
>
> If you are using uimaFIT, you can use the AggregateBuilder, e.g.
>
> builder = new AggregateBuilder();
> builder.add(analysisEngineDescription1, "output", "data");
> builder.add(analysisEngineDescription1, CAS.NAME_DEFAULT_SOFA, "output");
> AnalysisEngineDescription aggregateEngine = builder.
> createAggregateDescription();
>
> Cheers,
>
> -- Richard
>
> On 28.08.2015, at 23:25, Matthew DeAngelis <ronin78@gmail.com> wrote:
>
> > Hi all,
> >
> > I have been hitting my head against this all afternoon, so I figured I'd
> > ask.
> >
> > I am using a CollectionReader to read in HTML files and then using an AE
> to
> > add a view that contains the "clean", plaintext versions of those files.
> I
> > would like to pass the plaintext version to an AE downstream in the
> > pipeline (Stanford NLP's tokenizer from DKPro, specifically, for now);
> > currently, it only sees and tokenizes the original HTML file. Since the
> > tokenizer is not written by me, I would prefer to be able to modify my
> > view-creating AE instead of mucking about in someone else's.
> >
> > I have tried using an Aggregate Analysis Engine to combine the AEs
> together
> > and Sofa mappings to connect the views, but the downstream AE does not
> have
> > any input views to connect to. When I add an input view named
> > "_InitialView" to both the Aggregate AE and the descriptor for the
> > tokenizer, I get an interesting-looking NPE in some deep sub-function of
> > the tokenizer that suggests I have not been successful.
> >
> > Is there a way to tell my AE what view to pass as the "_InitialView", or
> > some way to set a different default view to pass?
> >
> >
> > Regards and thanks,
> > Matt
>
>

--089e01493fa82d0d64051e704085--