Derived Events and States (was Re: [RSP] Call Minutes)

Darko,

Regarding this text on complex events:

As stated in [1]: "Simple events happen at a single time instant, complex events start and end happening whereas facts always have a time interval during which they are valid.".

I commented this on the mailing list before, and I also added some comments in September onto the discussion page of the document you reference. Please check:
http://www.w3.org/community/rsp/wiki/Talk:RDF_Stream_Models

Once more, a complex event is - as defined by David Luckham and others who started to use the term - "an event that summarizes, represents, or denotes a set of other events" [Event Processing Glossary v. 2<https://www.google.fi/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&ved=0CEEQFjAB&url=http://www.complexevents.com/wp-content/uploads/2011/08/EPTS_Event_Processing_Glossary_v2.pdf&ei=pYOrUtykJsih4gSvuIGQCA&usg=AFQjCNHAfEs0uJN-MKwUF9bYN4zhBd0HLA&sig2=syJlb_7p2SzD3427JBi8mQ&bvm=bv.57967247,d.bGE>].

As an example, let us define complex event "D", which is triggered when events "A", "B" and "C" happen within one minute. Once we have a time window of one minute containing A, B and C, D will be synthesized at that instant. D doesn't "start happening" with the first one of the triggering events. And it certainly doesn't "end happening" with the last one of them, quite the opposite.

We recognized already with the previous email discussion that the definition of complex events has problems for us (ref. my email to the list Oct 2nd): the whole Event Processing Glossary is only talking about events; it doesn't even mention states (a.k.a temporally valid facts a.k.a situations) or static facts (a.k.a slowly evolving knowledge a.k.a background information). Whereas I'm pretty sure that we would like our "complex events" to accept a combination of everything - events, states and static facts - as triggers. Therefore I tried to start working on something that we could use, and the current draft of that is in the "Concepts and Definitions<http://www.w3.org/community/rsp/wiki/Concepts_and_Definitions>" document in the wiki, available for everyone to improve on:

-----------
"Derived event": An event that summarizes, represents, or denotes a combination of zero or more events, temporally valid facts or static facts such that there is always at least one triggering event object or temporal fact to establish the time of event occurrence.
-----------

It is overloading the term "derived event", because in the EPGv2 it is defined as "an event that is generated as a result of applying a method or process to one or more other events", but at least it is a little bit less used than "complex event". Better proposals are very welcome.

Coming back to the subject of ETALIS and intervals, I understand that combining intervals (situations) using operators like AND and OR will yield intervals as answers, but those combinations of temporally valid facts are not complex events, but something else. One initial attempt has been brewing in the Concepts and Definitions document:

-----------
"Derived state": A state that summarizes, represents or denotes a combination of one or more states, zero or more static facts and zero or more events.
-----------

Summarizing, I strongly -1 statements like "complex events start and end happening", because they don't seem to make sense. Otherwise I'm happy to see support for proper handling of temporally valid facts.

BR,

Mikko



On 13. Dec 2013, at 6:12 PM, Anicic, Darko wrote:

Axel,

-----Ursprüngliche Nachricht-----
Von: Axel Polleres [mailto:droxel@gmail.com] Im Auftrag von Axel Polleres
Gesendet: Donnerstag, 12. Dezember 2013 18:04
An: Anicic, Darko
Cc: Emanuele Della Valle; Jean-Paul; public-rsp@w3.org<mailto:public-rsp@w3.org>
Betreff: Re: [RSP] Call Minutes

Darko,

you are bringing validity time into the picture, it seems. whereas I
understood that the use for time stamps at the moment was mainly meant for
transaction time. [1] is that right? If we have use cases that need
validity time, I also think intervals make sense (FWIW, we have designed an
extension for SPARQL to query validity time intervals [2]).

Yes, for validity time of facts, but also for defining the time on which complex events have been derived. As stated in [1]: "Simple events happen at a single time instant, complex events start and end happening whereas facts always have a time interval during which they are valid.". Hence simple events could be represented as a graph with a single timestamp, complex events and facts could be represented as graphs defined on intervals. The meaning of the intervals for complex events and facts have different interpretations (and need semantics to be defined), i.e., complex events have been derived (occurred in the past) in an interval, whereas facts are valid in an interval (an interval possibly extended until certain point in the future).

Essentially if we have the interval-based graphs (e.g., with startTime and endTime where the startTime is optional), then we can represent the all three categories with no overhead.

Cheers,
Darko

[1] http://www.w3.org/community/rsp/wiki/RDF_Stream_Models#Temporal_Graphs


best,
Axel

1. http://en.wikipedia.org/wiki/Temporal_database
2. http://www.sciencedirect.com/science/article/pii/S1570826811000771

--
Prof. Dr. Axel Polleres
Institute for Information Business, WU Vienna
url: http://www.polleres.net/  twitter: @AxelPolleres

On Dec 12, 2013, at 5:05 PM, "Anicic, Darko" <darko.anicic@siemens.com<mailto:darko.anicic@siemens.com>>
wrote:

Hi Jean-Paul, Emanuele and all,

I like the proposal too, but I would not limit the proposal only to one
time stamp. As we discussed before, I find the interval-based definition
better, e.g., with startTime and endTime. The startTime may be optional in
cases where the point-based timestamp is preferred, e.g., similarly as in
[1] - a related approach to the Streaming Linked Data Framework, sent by
Emanuele (see #6 in Section Vocabulary from [1]).

Cheers,
Darko

[1]   http://km.aifb.kit.edu/sites/lodstream/

Von: Emanuele Della Valle [mailto:emanuele.dellavalle@polimi.it]
Gesendet: Dienstag, 10. Dezember 2013 11:15
An: Jean-Paul; Anicic, Darko; public-rsp@w3.org<mailto:public-rsp@w3.org>
Betreff: Re: [RSP] Call Minutes

Hi Jean-Paul, Darko and all,

I like the proposal. You may want to check out [1] where a similar
approach was proposed. We named the graphs o *instantaneous graphs*
(shortly iGraphs) and the graph containing the meta data Stream Graph
(shortly sGraph). This solution is currently implemented in our Streaming
Linked Data Framework [2].

GRAPH :sGraphA {
 :iGraph-1 :timeStamp "t_1"^^xsd:dateTimestamp .
 :iGraph-2 :timeStamp "t_2"^^xsd:dateTimestamp .
 .
 :iGraph-n :timeStamp "t_n"^^xsd:dateTimestamp .
}

GRAPH :iGraph-1 {.}
GRAPH :iGraph-2 {.}

GRAPH :iGraph-n {.}

Cheers,

Emanuele

[1] http://ceur-ws.org/Vol-628/ldow2010_paper11.pdf
[2] http://disi.unitn.it/~themis/publications/iswc13.pdf



On Dec 9, 2013, at 5:07 PM, Jean-Paul
<jpcalbimonte@delicias.dia.fi.upm.es<mailto:jpcalbimonte@delicias.dia.fi.upm.es>>
wrote:


Hi,

The one in [1] is a very generic model proposed by Axel. We agreed to
start from it to define the RDF Stream model.
If you see the minutes, you will see we are pointing to something that
*might* look like this:

o :timeStamp "foo"^^xsd:dateTimestamp . GRAPH o { t1, tn}

We have set an ACTION, that consists in formalizing these ideas in the
wiki. It is our homework for next call. But we can surely discuss here as
well.

best regards,
jp





2013/12/9 Anicic, Darko <darko.anicic@siemens.com<mailto:darko.anicic@siemens.com>> Hi all,

since I was not present at the last telco, could you please let me know
whether the RDF stream model that was discussed actually is the one
described as "graph oriented" in [1], or there are some differences in the
two?
Cheers,
Darko

[1] http://www.w3.org/community/rsp/wiki/RDF_Stream_Models

Von: jean.paul.ik@gmail.com<mailto:jean.paul.ik@gmail.com> [mailto:jean.paul.ik@gmail.com] Im Auftrag
vonJean-Paul
Gesendet: Samstag, 7. Dezember 2013 00:19
An: public-rsp@w3.org<mailto:public-rsp@w3.org>
Betreff: [RSP] Call Minutes

Hi all,

Please find in the wiki the minutes of today's call. Next one is on Dec
20th, and is the last for 2013. We will restart in January, the 17th.

http://www.w3.org/community/rsp/wiki/Telecon_06.12.2013

Thanks again for your support.

Good weekend!
Jean-Paul

--
jpcik!



---------------------------------
I have moved to EPFL, please email me to:
jean-paul.calbimonte@epfl.ch<mailto:jean-paul.calbimonte@epfl.ch>



--
jpcik!