What is an assertion in the DIKB?


The question has arisen in developing some definitions for DIDEO what is an ‘assertion’. I think we have this worked out already but maybe need a reminder – specifically, assertions are those entities that MP:formalizes MP:Claim entities as described in our ISWC paper. These are NP:Assertion individuals and are used in full Nanopublications as described in the paper and shown in numerous examples generated with the DIKB-Micropublication code base:






I might be wrong, but it seems to me that based on the paper the assertion as such does not need to be represented in the schema underlying the knowledge base.



Don’t NP:Assertion and MP:Claim need to be in the schema?



For the evidence base, yes. Mathias is saying that it’s not necessary for the knowledge base. My recollection is that we extract from the evidence base those assertions with a “certain level” of support into the KB, but pulling over all the evidence for them is not necessary.

I suppose if there is to be some link back to the evidence base, however, we might need some way to do that.



@Bill_Hogan has the evidence base / knowledge base distinction correct. There is a link back to the evidence base through the NP:Assertion instances using is MP:Formalizes. So, walking through an example:

  1. the nanopub in the knowledge base for the PDDI “diltiazem - triazolam”: http://dbmi-icode-01.dbmi.pitt.edu/describe/?url=http%3A%2F%2Fdbmi-icode-01.dbmi.pitt.edu%2Fmp%2Fddi-spl-annotation-nanopub-1

  2. The NP:Assertion component of the nanopub: http://dbmi-icode-01.dbmi.pitt.edu/describe/?url=http%3A%2F%2Fdbmi-icode-01.dbmi.pitt.edu%2Fmp%2Fddi-spl-annotation-np-assertion-1

  3. The MP that the assertion formalizes: http://dbmi-icode-01.dbmi.pitt.edu/describe/?url=http%3A%2F%2Fdbmi-icode-01.dbmi.pitt.edu%2Fmp%2Fddi-spl-annotation-claim-1

The nanopubs were created first using the evidence base MP and NP:Assertion RDF as input to an OWL reasoner to infer PDDI interaction statements by reasonable extraction. Observed and inferred PDDIs were then assembled into a KB of full Nanopublications after applying SPARQL to filter MP claims based on certain evidence types, and then assembling the remaining NP:Assertions into nanopubs with attribution etc.



Fine. Then the point is that we have TWO schemas – one for the knowledge base and one for the evidence base.
But the KB and EB are interlinked as @rdb20 points out above.



Based on this post, it seems that mp:methods will get formalized within the evidence base.

Shouldn’t we formalize the mp:claims within the evidence base as well?

In that case, both the evidence base and knowledge base would be formalized.
Then the distinction between the knowledge base and the evidence base is in the level of belief – not the level of formalization.

Thoughts, @rdb20, @mbrochhausen, @Bill_Hogan ?



I’m probably not clear on how to detect the boundaries between the EB and KB when looking at the links. The MP:claim at this link is formalized as the NP:assertion already: http://dbmi-icode-01.dbmi.pitt.edu/describe/?url=http%3A%2F%2Fdbmi-icode-01.dbmi.pitt.edu%2Fmp%2Fddi-spl-annotation-claim-1

Whether that formalization occurs in the EB before the NP is moved to the KB, or occurs only in the KB, I can’t tell. But that’s potentially a good thing: if someone wants more info, he or she is seamlessly linked back to the EB.