Re: Abstract Core Ontology for SWSL Processes

At 10:51 AM 24/01/2004 -0800, David Martin wrote:
>Will the representation of the temporal aspects of services get used for 
>practical purposes on the Semantic Web?


I would say very much so. Off the top of my head and in no particular order 
let me try a few...

a) For estimating min/max run times for some multi-service processes.

b ) For sequencing and making reservations on and contracts for important 
services as their time in a process gets close in a specific enactment 
instance.

c) For specifying con currency requirements.

d) For specifying required delays and cooling off periods in processes - 
say involving humans in the loop)

e) I would expect resource and cost calculations to be used along side 
these as well.. including the interaction of resource usage and time where 
cheaper rates might be available for some services at quiet times.  I can 
see some aspects of this and QoS related links to loading in the work at 
Imperial College on ICENI for example.


>If so, what kinds of questions will these representations be used to 
>answer?  What are the most important questions for them to be able to answer?


Ah, you can assume that will always be changing.  I am not a believer in 
"design one now and it will be right".  We need an approach that allows 
everything from trivial representations to VERY complex and unforeseen 
ones.  Hence my proposed abstract extendible core model for SWSL.

Our work involves large scale planning and scheduling so we need to process 
fast... so we have few temporal logic type temporal constrains in most 
domains - though our approach allows for this with suitable plug in 
constraint managers. More usually we use a simple representation of time 
points (with 2 being associated with each begin and end of any activity or 
process node) with "before" constraints on which you can put min/max 
bounds.  We actually use a triple min/estimate/max in fact.  We can have 
the before relation then on begin and end points of any activity, step or 
overall process (including dummy nodes to make the orderings 
simpler).  Very PERT chart style.  But it does allow straightforwardly for 
"ladder activities"  begin -> begin orderings, concurrency (min/max=0) 
specifications, (end -> end orderings  and concurrent end requirements (end 
-> end with min/max=0), etc.

Others will wish to use temporal logics such as James Allen's interval/13 
relations work.

I see no reason to assume one representation for temporal constraints in 
SWSL. So long as the ones used is described and reasoners provided for it.

>Are there any systems in widespread use that apply temporal reasoning to 
>solve practical problems (and which might have some relevance for SW services)?

Yes, we (and others) have deployed AI planners and schedulers that employ 
quite sophisticated temporal constraint management on large scale problems 
- typically orders of magnitude larger than will be the norm in SWS - 
including examples with 10,000 time points and their relationships.  Some 
of these algorithms are based on incremental versions of OR algorithms like 
linear programming methods.

There are temporal constraint checkers and systems that are used embedded 
in planners, etc.  Early examples were the time map management systems and 
the various families of those.  Even large corporate versions like GE 
Tachyon and Honeywell's TMM appeared and have been used outside if their 
labs. Both can be found through Google.  Pat may be more up to date than me 
on the modern examples of these.

Cheers, Austin

Received on Sunday, 25 January 2004 06:29:42 UTC