On 7/12/2012 2:47 AM, [email protected] wrote: > Dear Marshall, > > On Tue, Jul 10, 2012 at 7:53 PM, Marshall Schor <[email protected]> wrote: > >> The document analyzer has a "feature" that only shows types that are >> declared to >> be "outputs" in the capabilities section of the descriptor that it is >> running. >> > Yes, we discovered that "feature". But thanks for mentioning that, I think > this is useful also for others. > > However, our question of how and whether non-Annotation-derived types are > visualized in the document analyzer is still not really answered.
I don't recall exactly (would need to read the source code...) but I think that the document analyzer is only set up to show feature structures that are annotations. You can use other tools, such as the CVD tool which I think has an ability to show all the feature structures, instead of the Document Analyzer. -Marshall > > We set both the mention type and the relation itself as output in > the capabilities page. > > However, if we run the document analyzer, the relation is only 'visible' > whenever we make Relation a subtype of Annotation and set the end/begin of > the relation to the end/begin of the mention, respectively. But this > obviously implies that for the relation the whole substring between the > two mentions is highlighted, and not just (as we would wish) the two > mentions themselves (as an annotation has only one end/begin). > On the contrary, if we don't set the end/begin of the relation, then it's > not shown at all, not even despite the fact that the relations' two > features (the two entities) have each begin/end properly set. In more > detail, the "relation" checkbox is shown, but by enabling it nothing is > 'highlighted' > The same holds if we make Relation a subclass of TOP directly, still with > two features (mention1 and mention2 which are Annotation types with proper > begin/end set). We don't see it in the document analzyer. > > Thus, is it the case that such 'abstract' types (with what I mean > non-Annotation-derived types and hence types without begin/end) are simply > not shown in the document analyzer? > (thus to 'see' them we have to use the debugger?) We'd simply like to find > out if this is the case or we do miss something. > > Here is an illustration from the UIMA which shows what we'd like to have: > http://uima.apache.org/d/uimaj-2.4.0/overview_and_setup.html#ugr.ovv.conceptual.aes_annotators_and_analysis_results > > Thanks a lot for your help. > > B. > > >> So - just add the types you wish to show to that. See >> >> http://uima.apache.org/d/uimaj-2.4.0/tutorials_and_users_guides.html#ugr.tug.aae.testing_your_annotator >> and scroll to the bottom of the section and read the "Note". >> >> -Marshall >> >> >> On 7/9/2012 8:43 AM, [email protected] wrote: >>> On Fri, Jul 6, 2012 at 2:57 PM, Burn Lewis <[email protected]> wrote: >>> >>>> We created a type system for a project that included entries for >> Relations >>>> and RelationMentions ... you may wish to check GaleEntityTypes.xml in >> the >>>> GALE Multi-Modal Example in the Apache UIMA Addons and Sandbox. >>>> >>>> Thanks for this hint. >>>> You may find it easier to view TOP-derived types such as Relation with >> the >>>> CAS Visual Debugger >>>> >>> But is it actually the case that the document Analyzer visualizes >>> only Annotation-derived types? >>> Thus any TOP-derived type (like Relation) will not be shown in the >> document >>> Analyzer, even though it might contain references to Annotation-derived >>> types (like the two entity mentions that contain begin/end tokens)? >>> >>> Hence, to actually visualize TOP-derived types I need to use the CAS >> visual >>> debugger, there's no way to have them in the document analyzer? >>> >>> Thanks. >>> B. >>> >>> >>>> ~Burn. >>>> >> >>
