I think it's pretty bad.  If you use the XML detagging feature of
RunAE or DocumentAnalyzer, and try to run a remote AE (or an aggregate
containing a remote AE), it generates invalid XCAS files that cannot
be read.

I've already this morning put in code that checks for Sofa mapping in
a remote and throws an exception (which was already an issue marked
for 2.1, though Eddie never got around to doing it), and I've updated
the documentation.

-Adam

On 1/22/07, Marshall Schor <[EMAIL PROTECTED]> wrote:
I'm not fully following all this, but is this a show-stopper that needs
fixing in 2.1?

-Marshall

Adam Lally (JIRA) wrote:
>     [ 
https://issues.apache.org/jira/browse/UIMA-213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12466142
 ]
>
> Adam Lally commented on UIMA-213:
> ---------------------------------
>
> Unfortunately I don't think it's possible to make this work with any Sofa 
mapping as it's currently implemented.  I thought I could map the CPE's 
_InitialView to the XmlDetagger's output Sofa (plainTextSofa), but that won't work 
since the XmlDetagger will get an error when it tries to create the _InitialView.
>
>
>> DocumentAnalyzer/RunAE tools don't support XML detagging and Remote Vinci AEs
>> -----------------------------------------------------------------------------
>>
>>                 Key: UIMA-213
>>                 URL: https://issues.apache.org/jira/browse/UIMA-213
>>             Project: UIMA
>>          Issue Type: Bug
>>          Components: Tools
>>            Reporter: Adam Lally
>>         Assigned To: Adam Lally
>>             Fix For: 2.1
>>
>>
>> It generates invalid XCASes as output if you try to do this.  This is 
because Sofa mapping is used to do XML detagging, but Sofa mapping isn't implemented 
correctly for Vinci services (in fact it seems completely broken now).  Anyway it is 
slated to be completely removed for v2.1 (see UIMA-122).
>> It may be possible to do Sofa mapping differently so that the detagged text 
is stored in the _InitialView, so that Sofa mapping isn't needed for the remote.  
Failing that, we'd just need to put in a better error message until such time as we 
implement Sofa mapping for remotes in a reliable way.
>>
>
>


Reply via email to