SBPOI
These are just notes, they need polishing.
POI Pattern
A Place Of Interest is a micro-scale location where at least one placed activity can happen that is of interest to a (significant) amount of people (and stable over time.)
Scuba diving place xample: field of water depth → ScubaDivingActivity
→ Constraints (>3m)→ Boundary (cookie cutter) → bounded region on the surface of the Earth → named places (for those of interest). (begs question of what is "micro-scale" -- I suggest we use something along the lines of "resolves to spatial resolution at a scale of high likelihood of clear visual contact, or unamplified auditory contact between two individuals -- this would at least give us a focus for now, we might later consider if this can even be generalized (i.e. scale independent)) (while "Activity" is probably the motivation behind many POI's, I think it will be stretching it for certain use cases-- e.g. POI's notable simply because they are outliers/anomalies or otherwis notable, e.g highest local elevation point, places where one can't reach; places where we want to avoid; these are interesting, but not motivating an "Activity" per se. Although, we could contort the concept of "Activity" to encompass these types of situations.) ToDo's
subtypes or instances or singletons (of activities) (stereotypes versus KB of individual enriched instances ) >> blank notes
which constraints
relations between place and activity
'place' or 'location'
at least one time stamp?
what means 'stable over time'
significant? amount of people, do we need at least two?
exercise of an OGC Feature alignment. How would a sub-classing of ogc.feature by place would affect feature typing (relation to sub-classes of activities)
The notion of point-of interest? Restricted to LBS-like concept? Aren't there places without activities? Potential activity is a kind of interest (does this include interest?) Activity types of activities: See Activity, Do Activity POI where object of activity is a POI (or something else)? [This paragraph is obsolete now, I'm just keeping it for now in case the argument may stil be needed.] A preliminary remark on possible OWL modeling related to the pattern. It seems to me (?) that we have the following constraint on "correct" instances of the pattern: If x is a place, which affords an activity y, which in turn defines a boundary z, then z must cut out x. Formulated as a first-order predicate logic formula/rule, this would be affords(x,y) AND defines(y,z) AND boundary(z,w) --> x = w. This is a formula which (in full generality) cannot be modeled in OWL. That's not really a problem for the pattern, though: This constraint could, for example, be (a) a constraint which is formulated in english (i.e. not in a formal way), or (b) formulated using first-order predicate logic, or (c) some other formalisms extending OWL (there are a couple of choices, I don't think it's really critical - but it's an interesting discussion at least for me). It looks relatively straightforward to me to model the rest of the pattern in OWL. [End of obsolete paragraph] Re. logical modeling: An activity defines a boundary, which in turn gives rise to possibly *several* places, which in turn afford the activity. In logical terms, we can express this as the (first order predicate logic) rule defines(a,b) AND cuts_out(b,p) --> affords(p,a) This can be expressed in OWL using property a chain and a property inverse. Note, however, that this means that "affords" is non-simple, i.e. its usage in OWL 2 DL has some restrictions (e.g., "affords" cannot appear in cardinality constraints). Another global restriction ("regularity") also has to be observed.These are probably unproblematic but need some thinking (if it turns out to be problematic, there should be workarounds). Another constraint: If activity x has a realization y, and at the same time x is directed at z, then y must be directed at z. In logical terms: directed_at(x,z) AND hasRealization(x,y) --> directedAt(y,z) this is expressible in OWL as a role chain (no problems with regularity).
