- From: Roland Stühmer <mail@roland-stuehmer.de>
- Date: Fri, 04 Apr 2014 10:51:53 +0200
- To: Abraham Bernstein <bernstein@ifi.uzh.ch>
- CC: "public-rsp@w3.org" <public-rsp@w3.org>
Abraham Bernstein wrote:
> A stream is a sequence (G, τ ) where τ ∈ T the set of all *intervals*.
>
> τ = [t_s, t_e], where t_s denotes the start time and t_e denotes the
> end time.
>
> As a short hand notation we could say that when t_s = t_e then one can
> only write one of them.
+1 for Intervals and +1 for the short notation (defaulting to time points)!
Darko and I are using intervals internally because it fixes problems
with instantaneous occurrences in complex event, see [1].
Contrary to an intuitive understanding the expressions E1->(E2->E3) and
E2->(E1->E3) ("->" is "followed by") are equal with point-based
semantics. The order of the events E1 and E2 in the example is
inconsequential because the whole event is detected if the event in
before the parentheses is detected before E3. No relation between events
in position one and two is specified. Both expressions are fulfilled by
the same event histories: e1,e2,e3 and e2,e1,e3.
This is not expected from the "followed by" relation. Interval-based
semantics solves these problems by viewing complex events as occurring
over the interval from the occurrence of the first constituent event,
the initiator, to the end of the last constituent event, the terminator.
Best!
Roland.
[1] Section 4 "Why Detection Conditions are Inadequate" of
http://users.ecs.soton.ac.uk/jca/dexa2002.pdf
Galton, A. & Augusto, J. C. (2002), Two Approaches to Event Definition,
in 'DEXA '02:
Received on Friday, 4 April 2014 08:52:17 UTC