Grounding Domain KBs

DCP forms domain knowledgebases in two directions, top down from user input and bottom up from system ontologies.

You specify the apps and data of your domains by conversational interaction with the DCP. This results in the normative text defining end use domains in natural language which is the reference form of them in this system. These are at a high level and are grounded in imprecise knowledge sources obtained from the users ranked by role and their experience in the real world and general linguistic culture distilled from the web by language models. No specific expertise other than as subject matter experts for the domains in question is required of the authoritative users. DCP performs the reduction of the reference form to data and knowledgebases and programming language text and operates them in the real time of the domain.

Beneath this purposive and normative knowledge basing of your domains there is formal and structured base of knowledge inherent in the system ontonlogy and taxonomy mechanism. These are intended to be worked by DevOps knowledge engineers. Currently, the main basing of this knowledge is a curated extraction from Wikipedia to populate a base Suggested Upper Merged Ontology (SUMO) as a base of well grounded knowlege which the system interpolates with the top level specification of end use domains.

Here is the order of precedence as far as DS Knowledge Sources are concerned.

1. Current system intrinsics.
2. Subject Matter (authoritative user) expertise.
3. Ambient linguistic attestations.

Both of these levels of the system knowledgebase are dynamic and the system can autonomously initiate changes, subject to authorization by a user with sufficient privilege, to the system level KBs and this can impact the end level domains.