This document discusses representing Parsimonious Covering Theory (PCT) in OWL-DL. PCT is an abductive logic framework that seeks to account for observed symptoms or manifestations with plausible explanatory disorders or hypotheses. The key steps are: (1) modeling disorders, manifestations, and their causal relationships in OWL, (2) representing observations as an "Explanation" class that has causes relationships to the observed manifestations, and (3) deducing valid explanatory disorders as instances of the Explanation class. Converting PCT to OWL in this way allows taking advantage of OWL reasoning to find minimal explanations over linked sensor data on the web.
2. OWL Experiences and Directions (OWLED 2011) Representation of Parsimonious Covering Theory in OWL-DL Cory Henson, KrishnaprasadThirunarayan, AmitSheth, Pascal Hitzler Ohio Center of Excellence in Knowledge-enabled Computing (Kno.e.sis) Wright State University, Dayton, Ohio, USA 2
3. Find a set of entities (in the world) that explain a given set of sensor observations 3
4. Characteristics of a Solution Handle incomplete information (graceful degradation) Minimize explanations with additional information (anti-monotonic) Reason over data on the Web (i.e., RDF on LOD) Scalable (tractable) 4
7. Parsimonious Covering Theory (PCT) Web Ontology Language (OWL) minimize explanations tractable degrade gracefully Web reasoning Convert PCT to OWL 7
8. Parsimonious Covering Theory Goal is to account for observed symptoms with plausible explanatory hypotheses (abductive logic) Driven by background knowledge modeled as a bipartite graph causally linking disorders to manifestations disorder manifestation causes m1 d1 m2 d2 m3 d3 m4 explanation observations YunPeng, James A. Reggia, "Abductive Inference Models for Diagnostic Problem-Solving" 8
9. PCT Parsimonious Cover coverage: an explanation is a cover if, for each observation, there is a causal relation from a disorder contained in the explanation to the observation parsimony: an explanation is parsimonious, or best, if it contains only a single disorder (single disorder assumption) 9
10.
11. disorders (D) for all d ∈ D, write d rdf:type Disorder ex: flu rdf:type Disorder cold rdf:type Disorder manifestations (M) for all m ∈ M, write m rdf:type Manifestation ex: fever rdf:type Manifestation headache rdf:type Manifestation … causes relations (C) for all (d, m) ∈ C, write d causes m ex: flu causes fever flu causes headache … PCT Background Knowledge in OWL disorder manifestation causes fever headache extreme exhaustion severe ache and pain flu mild ache and pain stuffy nose sneezing cold sore throat severe cough mild cough 11
12. observations (Γ) for mi∈ Γ, i =1 … n, write Explanation owl:equivalentClass causes value m1 and … causes value mn ex: Explanation owl:equivalentClass causes value sneezing and causes value sore-throat causes value mild-cough explanation (Δ) Δrdf:type Explanation, is deduced ex: cold rdf:type Explanation flu rdf:type Explanation PCT Observations and Explanations in OWL and 12
13. Ohio Center of Excellence on Knowledge-Enabled Computing (Kno.e.sis) thank you, and please visit us at http://semantic-sensor-web.com Knoesis – Ohio Center of Excellence in Knowledge-enabled Computing Wright State University, Dayton, Ohio, USA 13