Re: Event significance

It is pleasing to read that such news organizations as Reuters (Misha) and 
Agence France-Presse (Laurent) are so much concerned  about such fundamental 
issues as: the nature of events (reported), assigning a significance to news 
events, and how this may be reflected in a machine-processed representation.

The passage below shows that Misha Wolf is generally suggesting the sensible 
plan of attack:

''A note on events vs event types: Though I use the term *event* in
the above, what I'm describing is really an *event type*.  A
hurricane (or a meteorological disaster) is an instance of an
event type, while Hurricane Katrina is an instance of an actual
event of that type.
We need to be able to "speak" of both events and event types.  It
is not valuable or even feasible to give every event an identity
(ie a URI) which would make it possible to make statements about it.
It is more plausible to give URIs to event types and to significant
events.''


Ontology comes before Semantics as Semantics comes before Syntax: there is 
an absolute priority of ontological axioms (or committments) over formal 
repesentations.
For the ontologies of events can be written down in a variety of formal 
languages and notations (XML, RDF, or what not); nevertheless, the essential 
information is not the form of the languages rather the content, the set of 
entity classes and types and individuals offered as a way of reasoning about 
the world. Put simply, the gist of the matter is notions like events and 
causal relations, and not whether we choose to write them as XML constructs, 
logical predicates or OWL constructs.

Though, we have to mention that there are some fundamental distinctions 
which may lie outside of the competency of the news agencies people. Above 
all, it is of crucial necessity to discriminate events as individual 
variables and changes as class variables; since
[an event (hurricane Katrina) is a change (windstorm, cyclone) happening 
(taking place, occurring or coming about) at (in, on near, or by) some place 
(New Orleans, Louisiana) over a period of time].

So the basic meaning and definition of the term 'event', 'occurrence', or 
'happening', its nature and content, lies in the notion of change (or 
''event type''); while its extensional (connotative) meaning may be found in 
the set of relationships that event has to other things. Then the 
significance of any individual event is defined by its possible causal 
consequences, impacts, or effects. This means that any natural event should 
be ascribed a [magnitude] quantity consisting of intensity (strength) and 
extent (reach) of actions. So, all sorts of disasterous events as 
catastrophes, calamities, cataclisms, or tragedies [meltdown, pestilence, 
deluge, plaque, tidal wave, tsunami, famine, or even apocalypse] pertain to 
the occurrences of enormous magnitudes that may or may not bring disaster 
and huge loss to humans, all depends on the quality of our predictive 
theories and intellectual technologies.

Bottom line: to adequately represent and reason about events, we have to 
trouble not about their formal representations and subjectively given 
meanings but rather about a consistent ontology of processes of changes.

Azamat Abdoullaev
EIS Encyclopedic Intelligent Systems LTD
http://www.eis.com.cy




----- Original Message ----- 
From: "Misha Wolf" <Misha.Wolf@reuters.com>
To: <newsml-2@yahoogroups.com>; <semantic-web@w3.org>
Sent: Saturday, March 11, 2006 9:13 PM
Subject: RE: Event significance



On 10 March, Misha wrote:

> Our journalists want to be able to indicate the significance of an
> event to one or more entities.  Consider hurricane Katrina.  It
> had a high significance to (among others) New Orleans, Louisiana,
> the oil industry.  So how could we indicate this in NewsML 2?

[...]

> So how can we design in significance?  My current thinking is that
> we:
>
> -  Add a significance attribute to the subject element.
>
> -  State that when this attribute is applied to a subject which
>    represents an entity, then the value of the attribute represents
>    the significance to that entity of any nearby subject which
>    represents an event.

A better approach would be to add a <significance/> child element
to the <subject/> element denoting the affected entity, with a link
of some kind to the event.  What kind of link?  There seem to be
two options:

1. An "idref" attribute linking to a <subject/> element denoting
   the event, eg:

   <subject code="iso3166-2:US-LA" relevance="80" confidence="80"
       created="2005-08-29">
     <sameAs code="usps:LA"/>
     <childOf code="iso3166-1:US"/>
     <title xml:lang="en">Louisiana</title>
     <title xml:lang="fr">Louisiane</title>
     <significance idref="foo">80</significance>
     <significance idref="bar">80</significance>
   </subject>

   <subject code="nc:03007000" type="typ:event" id="foo">
     <childOf code="nc:03000000"/>
     <title xml:lang="en">Meteorological disaster</title>
     <title xml:lang="fr">Désastre météorologique</title>
   </subject>

   <subject code="nc:03005000" type="typ:event" id="bar">
     <childOf code="nc:03000000"/>
     <title xml:lang="en">Flood</title>
     <title xml:lang="fr">Inondation</title>
   </subject>

2. A "code" or "about" attribute holding the CURIE of the event, eg:

   <subject code="iso3166-2:US-LA" relevance="80" confidence="80"
       created="2005-08-29">
     <sameAs code="usps:LA"/>
     <childOf code="iso3166-1:US"/>
     <title xml:lang="en">Louisiana</title>
     <title xml:lang="fr">Louisiane</title>
     <significance code="nc:03007000">80</significance>
     <significance code="nc:03005000">80</significance>
   </subject>

   <subject code="nc:03007000" type="typ:event">
     <childOf code="nc:03000000"/>
     <title xml:lang="en">Meteorological disaster</title>
     <title xml:lang="fr">Désastre météorologique</title>
   </subject>

   <subject code="nc:03005000" type="typ:event">
     <childOf code="nc:03000000"/>
     <title xml:lang="en">Flood</title>
     <title xml:lang="fr">Inondation</title>
   </subject>

I'm inclined towards the second option.

A note on events vs event types: Though I use the term *event* in
the above, what I'm describing is really an *event type*.  A
hurricane (or a meteorological disaster) is an instance of an
event type, while Hurricane Katrina is an instance of an actual
event of that type.

We need to be able to "speak" of both events and event types.  It
is not valuable or even feasible to give every event an identity
(ie a URI) which would make it possible to make statements about it.
It is more plausible to give URIs to event types and to significant
events.

Misha
------------------- NewsML 2 resources ------------------------------
http://www.iptc.org/        | http://www.iptc.org/NAR/
http://www.iptc.org/NAR/1.0 | http://groups.yahoo.com/group/newsml-2/


To find out more about Reuters visit www.about.reuters.com

Any views expressed in this message are those of the individual sender, 
except where the sender specifically states them to be the views of Reuters 
Ltd.

Received on Monday, 13 March 2006 21:13:06 UTC