Web annotations and NLP


(zachary.whitley) #1

I just wanted to throw an idea for Stardog out there. I doubt I’ll have time to work on it independently any time soon. It would be really cool if Stardog supported Web Annotations. The basic idea would be that if you could generate new NLP models from the annotations you could support a really nice NLP work flow by providing an initial set of annotations, build a model and generate new annotations from the model, edit the new annotations, rinse and repeat. Record all of it using VCS so you have a nice history of how the model was built maybe even build different models based on who did the annotation maybe select the users with the highest agreement across annotations with the help of some Stardog machine learning. Integrate the Brat rapid annotation tool into StardogStudio and that would be one sweet NLP toolkit.

I should probably add that obviously Stardog trivially supports Web Annotations because it’s just triples but I was more along the lines of supporting NLP model generation and output. NIF is interesting but it might be a bit heavy weight but maybe it can be lightened up with the use of reasoning.