[jira] [Commented] (UIMA-3661) Ruta: specification for "Filter & Retain"

["Peter Klügl (JIRA)" <de...@uima.apache.org>]

    [ https://issues.apache.org/jira/browse/UIMA-3661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13915955#comment-13915955 ] 

Peter Klügl commented on UIMA-3661:
-----------------------------------

Yes, that is a stupid problem. Just to clarify it for others: The annotation with the covered text "XY (KQ)" is not matched, because parenthesis are not visible. One would think that only SegmentElements starting with a parenthesis are not matchable. This is quite counterintuitive, but the begin AND the end need to be handled the same way because of the variable matching direction supported by Ruta.

I agree that we need a better and cleaner concept for visibility. I always thought that our concept is much more powerful than related implementations such as JAPE. They are type-based, we are coverage-based, but that comes with a price.

I invite everyone to discuss what should be visible/matchable.

> Ruta: specification for "Filter & Retain"
> -----------------------------------------
>
>                 Key: UIMA-3661
>                 URL: https://issues.apache.org/jira/browse/UIMA-3661
>             Project: UIMA
>          Issue Type: Wish
>          Components: ruta
>    Affects Versions: 2.1.0ruta
>            Reporter: Martin Toepfer
>
> It would be nice if we had a specification for inference with filtered and retained types, as well as some options for handling filtered/retained types (covered text vs. annotations).
> For instance, I am trying to merge SegmentElement annotations into Segment annotations (Segment is roughly SegmentElement+). I am not interested in parenthesis or colons, so I add them to the filtered types. However, I have SegmentElements that contain parenthesis, thus, I add SegmentElements to retained types. Nevertheless, not all SegmentElements are used, because some SegmentElements end with closing parenthesis, e.g., "XY (KQ)", which is a valid element in my case.
> At the moment, I cannot use filter/retain for this problem. Changes may affect many parts of the inference, however, I think that it would be a very nice feature for UIMA Ruta 3.0.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)