Community:Situation classification - Revision history

EvaBlomqvist at 15:37, 12 October 2010

2010-10-12T15:37:31Z

AldoGangemi at 15:12, 12 October 2010

2010-10-12T15:12:04Z

AldoGangemi: WOP2010

2010-10-12T15:10:39Z

WOP2010

New page

{{TitleDescription Template
|Title=Situation classification
|Description=Classical problem-solving in AI and knowledge engineering requires a full-fledged knowledge representation system, including languages for representing the domain and the axioms/rules that hold for it (the problem space), as well as algorithms to find a solution to that problem, if any. Classical approaches use different languages for the representation of the problem versus the representation of the solution, e.g. declarative vs. procedural.
However, there are many requirements in domain modelling which require to talk about both the problem and solution spaces within a same universe of discourse. This is the case of legal knowledge (factual knowledge and legal cases vs. normative knowledge), services, planning and control knowledge (actual facts vs. expected facts), diagnostic knowledge and situational awareness (e.g. bare facts vs. typically unwanted facts), etc. All those cases fit into the abstract task of "classifying a situation" according to possibly incomplete, alternative or loose constraints, where those constraints must be explicit and explicitly linked to the representation of a situation.
There are content patterns for representing situation classification, such as descriptionandsituation.owl, which relies on reification of conepts and relations. These patterns can represent situations and descriptions, and their given links (e.g. entities of a situation that play roles from a description, values of a situation that fit parameters from a description, etc.).
But it's very hard to represent in general (not based on locally defined axioms, e.g. owl:equivalentClass axioms for a particular situation/description pair) how to:

1) make a situation emerge out of scattered facts
2) decide if a situation (or a set of facts) can be (partly or fully) classified under a description
3) evaluate which description fits best a certain set of facts
etc.

Classical approaches represent situation classification within some algorithm or in the semantics of some appropriate meta-model. The modeling issue here is:

---) can we approximate generalized representations, for at least some situation classification tasks, by using OWL2 (e.g. with features such as punning, property chains, reflexivity) and regular description logic reasoners. In other words, is it possible to represent situation classification so that it reduces to a concept classification problem?
}}
{{Graphical representation}}
{{Modeling Issue Template
|CompetencyQuestion=What situations satisfy a certain description? What descriptions can be (partly) satisfied by that situation? What situations (partly) satisfying a certain description can emerge out of this dataset?
|RelatedPattern=Description, DescriptionAndSituation, Situation, Classification, Parameter
}}
{{Additional information header}}

{{My references}}
{{Modeling Issue toolbar}}

@@ Line 17: / Line 17: @@
 {{Graphical representation}}
 {{Modeling Issue Template
 |Author=AldoGangemi,
 |Domain=General, Planning
 |CompetencyQuestion=What situations satisfy a certain description? What descriptions can be (partly) satisfied by that situation? What situations (partly) satisfying a certain description can emerge out of this dataset?
@@ Line 26: / Line 26: @@
 {{My references}}
 {{Modeling Issue toolbar}}