Hello everyone,

I have some problems to automatically extend the KIM Knowledge Base.
But perhaps I start to describe what I want to do and what I have done 
so far:

I am creating an client application accessing the KIM Api
to achieve automatic semantic annotation for a specific domain.
For example I try to identify open source projects on web-sites like 
sourceforge.net etc.

My progress so far:

I extended the PROTON ontology with some of my domain specific classes 
(e.g. OpenSourceProject as a subclass of protont/Project).
I also extended the Knowledge Base with some pre-defined Entities (e.g. 
Eclipse with some aliases) from a trusted source, until here the KIM 
annotation process works fine.

Now I want my application to find new (OpenSourceProject) instances 
automatically to extend the KIM knowledge base. For testing purposes I 
created some dummy JAPE rules in a new *.jape file and added this to the 
default main.jape file in my current context.

Running the KIM Client PR in GATE without the Instance Generator my 
domain specific annotations (e.g. OpenSourceProject) are correctly shown 
in the 'Annotation Sets' display in addition to the other annotations 
(like the internal annotations Lookup, split, unknown etc.). My JAPE 
rules also work with default ANNIE processing resources.

But I am having problems running the KIM client PR with the Instance 
Generator (my intention is to put my new detected annotations in the 
new-temp-triples.nt file). When running this PR in GATE the internal 
annotations disappear (which I think is OK) BUT ALSO my newly created 
ones disappear. As a consequence I have no desired entries in the *.nt 
file (Only 'default' entries like Organisation instances).

I do not understand this behaviour (of the Instance Generator). As it is 
mentioned in the KIM manual section 'Extending/Changing the Information 
Extraction Module(s)' I can edit the grammars for specific recognition
(but I do not understand why this is not really recommended to reuse the 
provided grammar and add specific ones).

I also tried to create a brand new IE Module. First I analysed the KIM 
'IE-no-keywords.gapp' file to identify the KIM 'Conditional Corpus 
Pipeline' components. But when I tried to rebuild it I had problems 
initialzing some KIM specific Processing Resources in GATE (e.g. a 
ClassCastException occured when I tried to create a KIM Gazetteer). So I 
also do not understand how to build a brand new IE Module for KIM.

Did I missed something to extend the kb? Thanks in advance!

Björn




_______________________________________________
NOTE: Please REPLY TO ALL to ensure that your reply reaches all members of this 
mailing list.

KIM-discussion mailing list
[email protected]
http://ontotext.com/mailman/listinfo/kim-discussion_ontotext.com

Reply via email to