Hi Steven, All, Firstly, let me note we recently released a new version of KIM which among other things features a SOAP Web service wrapper of most of the KIM functionality. You can check out the release notes at http://ontotext.com/kim/release-notes.html and the new documentation at http://ontotext.com/kim/doc/sys-doc/HomePage.html
So there are two parts in your question. It's easy to fix a bad classification in a single document. The API exposes all necessary methods to create a new correctly classified entity, loop over all annotations in a document, change their class and sync the document back into the document repository. Teaching KIM not to make the same mistake again is not yet supported in the API, because we haven't developed a workflow for the user to feedback information. Internally, we have a few avenues for controlling the output of the information extraction: ignore lists for entities that are frequently misclassified in a specific context, priority lists and additional term specification. All of those give us and you options to better configure KIM for a specific domain which may be news from UK, technical documents or, like in our latest use-cases, pharmaceutical documents from our partners at AstraZeneca and tourism advertisements. In particular, priority lists give the option to specify that which competing classification over entities has priority over the others. Additional term specification makes it possible for KIM to receive feedback like "Carling is not simply a http://ontotext.com/...#Product but http://domain-specific-ontology#Lager." I believe, that together with you and our community, we can design a workflow for user corrections of the annotations of a document. That workflow will lead to an API for corrections. The exact implementation of the API will probably be specific to the domain in which KIM is used. I'm looking forward to your input on that issue. Best regards, Marin Nozhchev Lead Developer, KIM Platform On Sun, 2008-07-20 at 12:05 +0100, steven shingler wrote: > Hi all, > > I have been looking at KIM - getting entities back from a web page. > > Sometimes, the entity may be classified incorrectly - for example - it > will say "X" is a location, whereas the user will know it is, in fact, > a person. > > Is there a way of "feeding back" a re-classification of that entity > type into the KIM ontology? - - So that it doesn't make the same > mistake next time? > > Any advice gratefully received, > > Many Thanks, > Steven > > _______________________________________________ > 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 _______________________________________________ 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
