W3C home > Mailing lists > Public > public-hcls-coi@w3.org > July to September 2008

FW: ER-EHR INTERESTING PAPER: Disaster Management, Ontology and the Semantic Web

From: Kashyap, Vipul <VKASHYAP1@PARTNERS.ORG>
Date: Mon, 21 Jul 2008 10:11:05 -0400
Message-ID: <DBA3C02EAD0DC14BBB667C345EE2D12402E8C73E@PHSXMB20.partners.org>
To: <public-hcls-coi@w3.org>
Cc: <public-semweb-lifesci@w3.org>
Interesting discussion at HITSP (Healthcare Information Technology Standards
Panel) chartered by ONCHIT (Office of the National Coordinator for Healthcare
Information Technology) in the US
 
---Vipul

________________________________

From: Stephen Hufnagel [mailto:SHufnagel@TIAG.NET] 
Sent: Monday, July 21, 2008 7:59 AM
To: HITSP-PROVIDER-PERSP-TC@MAILLIST.ANSI.ORG
Subject: ER-EHR INTERESTING PAPER: Disaster Management, Ontology and the
Semantic Web



I added this paper as appendix D of our ER-EHR status report. It  seems to
summarize our projects well.

from

http://polaris.gseis.ucla.edu/qhallida/DisasterManagementFinal.html

 

Disaster Management, Ontology and the Semantic Web

Quentin Halliday

June 14, 2007

Revised June 27, 2007

Introduction:

Recent disasters have spurred the search for technological systems that are
designed to aid the management of such calamitous events.  In the USA the
terrorist attacks of 9/11 and the devastation caused by hurricane Katrina in New
Orleans have been used as motivating factors for the exploration of the
possibilities of developing common standards that can be employed in the
aftermath of these emergencies.  The Asian earthquake and subsequent tsunami of
late 2004 also engendered efforts to design technological means of managing the
resultant loss of life and destruction of property.  At the same time, emergency
and disaster response agencies have been working on ways to improve the
management of many types of events
<http://www.gdrc.org/uem/disasters/1-what_is.html> , such as natural disasters
(including floods, droughts, earthquakes, cyclones, hurricanes, tornadoes,
typhoons, landslides and volcanic eruptions) and  man-made disasters (including
chemical accidents, vehicular collisions, oil spills, radiological accidents,
conflicts/wars, mass population displacement or refugee emigration, forest fire
and terrorist attacks).   

Dealing with these emergencies, which vary in size, effect and nature, involves
many types of agencies.  For example, even a small local event such as a car
crash will usually involve police, medical response, hospitals and highway
authorities and the heterogeneity of these bodies has given rise to the effort
to coordinate their responses through common descriptions of the event that can
be interpreted in the light of their individual expertise.  

After a brief description of the field of emergency management, this paper will
address the results of the efforts to standardize, paying particular attention
to the standards developed in the USA that are mostly taxonomic in nature.
After a description of the taxonomy technology, its architecture and its
deployment, there will be an examination of the rhetorical, political and
institutional setting of these developments to show how these projects are
maintained and coordinated.  This will lead into the vision that the instigators
of this work have for the future of their efforts, including the development of
ontologies, and this will be contrasted with other realizations of systems to
manage disasters.

The Nature of Emergencies:

The field of emergency or disaster management usually divides any situation into
three or four components <http://www.gdrc.org/uem/disasters/1-dm_cycle.html>
that roughly correspond to the before, during and after phases of any particular
event.  The Mitigation Phase is concerned with lessening the effects of possible
disasters by mandating policies that will minimize the effects of disaster
should one occur.  The implementation of building codes that regulate
construction in an earthquake zone is an example of this phase.  The
Preparedness Phase focuses on plans, exercises and warnings, thus putting in
place the mechanisms that will be employed, for example, to alert the population
and manage their orderly evacuation.  The subsequent Response Phase deals
directly with the results of the disaster and is aimed at mobilizing people and
materiel to support such functions as search and rescue and emergency relief.
Finally, the Recovery Phase involves the effort to return the affected
communities to normal, an example being the widespread use of FEMA trailers
after Hurricane Katrina that provide residents with temporary housing while they
rebuild.

The presence of a relatively settled consensus among participants of what
emergency management consists of means that the possibility for ontological
description exists.  However, taxonomic and ontological approaches to disasters
have concentrated almost exclusively on the Response Phase, dealing with the
immediate aftermath of an emergency including the mobilization of relief
agencies, the delivery of aid and the provision of emergency medical care. It is
possible to speculate that this is due to the well-publicized failures of
emergency services in this phase, such as the radio communication problems at
the World Trade Center or the widely reported but probably untrue happenings in
the Louisiana Superdome
<http://seattletimes.nwsource.com/html/nationworld/2002520986_katmyth26.html> .
While it is not possible to say whether this is the cause of the concentration
on the Response Phase, it is clear that other approaches could be taken: in the
commercial world, for example, business continuity plans
<http://www.yourwindow.to/business-continuity/bcp7.htm>  place at least as much,
if not more, emphasis on the Recovery Phase.

XML Technology in Disasters and Emergencies: What Has Been Agreed. 

At present the operational systems that employ common descriptions of events and
resources concerning disasters in the USA are based on XML and the first
component of the OASIS <http://www.oasis-open.org/home/index.php> -sponsored
project to develop common standards for disasters and emergencies was the Common
Alerting Protocol
<http://pdfdownload.bofd.net/pdf2html.php?url=http%3A%2F%2Fwww.oasis-open.org%2F
committees%2Fdownload.php%2F15135%2Femergency-CAPv1.1-Corrected_DOM.pdf>  (CAP).
The CAP is an OASIS standard adopted September 30, 2005.  Version 1.0
<http://pdfdownload.bofd.net/pdf2html.php?url=http%3A%2F%2Fwww.oasis-open.org%2F
committees%2Fdownload.php%2F6334%2Foasis-200402-cap-core-1.0.pdf>  was
superseded by version 1.1
<http://pdfdownload.bofd.net/pdf2html.php?url=http%3A%2F%2Fwww.oasis-open.org%2F
committees%2Fdownload.php%2F14759%2Femergency-CAPv1.1.pdf>  in October 2005.
The purpose of the CAP is to be a simple, general format for exchanging alerts
and warnings over all kinds of networks.  For this reason, the standard was
developed in XML as a "transport agnostic" means of permitting a consistent
warning message to be disseminated simultaneously over different systems.  The
aim is to simplify the alerting process while increasing its effectiveness.  The
CAP has been accepted by several major bodies in the US disaster management
field including the National Weather Service <http://www.weather.gov/alerts/>
and the US Geological Survey <http://www.usgs.gov/hazards/> .  The FCC also
recently issued an order
<http://pdfdownload.tsone.info/pdf2html.php?url=http%3A%2F%2Fwww.eic.org%2Fdocs%
2FFCCCAPEASnotice.pdf>  requiring users of the Emergency Alert System to adopt
CAP when FEMA adopts it.

CAP functions as a standalone format but can also be used as a payload in the
Emergency Data Exchange Language
<http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=emergency#technical>
(EDXL) that is now the focus of most attention among institutions that are
developing taxonomic approaches to disaster management.  This language is being
built piecemeal via the offices of OASIS and at present comprises three
components.  The first of these, the Distribution Element
<http://pdfdownload.tsone.info/pdf2html.php?url=http%3A%2F%2Fdocs.oasis-open.org
%2Femergency%2Fedxl-de%2Fv1.0%2FEDXL-DE_Spec_v1.0.pdf>  (EDXL-DE), is an OASIS
standard approved in June 2006.  The Distribution Element functions as a
container for other messages, providing the information to route payloads, such
as CAP, by specifying such things as agency type, geography, incident type and
the identification of sender and recipient.  The second component of EDXL, the
Hospital AVailability Exchange
<http://pdfdownload.bofd.net/pdf2html.php?url=http%3A%2F%2Fdocs.oasis-open.org%2
Femergency%2Fedxl-have%2Femergency_edxl_have-1.0-spec-pr02-1.pdf>  (EDXL-HAVE)
completed its period of public review in January 2007 and is in the final stages
of ratification as a standard by OASIS.  As its name suggests, HAVE is used to
communicate the status of a hospital and its resources to agencies that includes
such elements as emergency department status, bed capacity and types of services
available.  The third component, Resource Message
<http://pdfdownload.04340.com/pdf2html.php?url=http%3A%2F%2Fdocs.oasis-open.org%
2Femergency%2Fedxl-rm%2Fv1.0%2Fcd01%2FEDXL-RM-SPEC-V1.0.pdf>  (EDXL-RM), extends
the idea of HAVE to cover many types of resources that might be needed during
the Response Phase of an emergency.  Each individual Resource Message is
designed to enable requests and orders for human resources, vehicles, equipment
and supplies from agency to agency.  It is intended to support
machine-to-machine communication in an effort to avoid multiple entries of the
same information.  It is hoped that using Resource Messages at the time of an
event, or several events, will help to manage scarce resources efficiently since
the status of all resources will be available simultaneously to many agencies.
The period of public commentary on EDXL-RM closed on 8 June, 2007 and it now
moves into the ratification stage at OASIS.

Public deployment of these standards is at present limited to CAP.  For example,
the feeds from various agencies are scraped by the Global Disaster Alert and
Coordination System (GDACS <http://www.gdacs.org/>  ) and presented as an
aggregation of different types of emergencies around the globe.  GDACS is
designed as a clearing-house for disasters, bringing together alerts, situation
reports and coordination of response.  (When viewed from an ontological
perspective, GDACS is also developing a de facto set of disaster-types
<http://www.gdacs.org/about/alertlevels.asp>  that includes their nature and
severity.)

It looks likely that EDXL will be employed shortly by the National Incident
Management System (NIM <http://www.fema.gov/emergency/nims/index.shtm> S) in
their deployment of a database that will be implemented locally by emergency
management organizations and distributed nationally via a network. The NIMS
Integration Center (NIC) has typed 120 resources
<http://www.fema.gov/emergency/nims/faq/rm.shtm>  needed in emergencies and
encourages emergency response organizations to type their resources according to
this nascent ontology. Soon NIMS will deploy the Incident Resource Inventory
System (IRIS) <http://www.fema.gov/emergency/nims/rm/iris.shtm> , a free
database application into which communities can enter their emergency response
resources typed according to the NIMS list. This a move towards standardization
of resource types to be made available in a human-readable form. IRIS will use
EDXL to communicate resource availability.

It is noticeable that CAP and EDXL are a growing set of standards for messages
that are being developed through OASIS. The approach here has been to work from
the bottom up, starting with a common standard for alerts (CAP), adding a
mechanism for distributing alerts (EDXL-DE), then addressing the availability of
one critical resource in the Response Phase (EDXL-HAVE) and finally widening the
standard to encompass many resources (EDXL-RM).  From the point of view of
eventual ontological description of disaster management it is possible that this
type of approach that concentrates on particular "islands of uniformity" through
a standardized process might be relatively successful in isolating common
elements that practitioners can agree on for their domain.  The standardized
process involved also ties the XML elements to the domain and its communities by
first asking emergency practitioners to develop their requirements, then passing
these to technology companies (the Emergency Interoperability Consortium
<http://www.eic.org/> ) and then to OASIS.  This type of approach contrasts with
a top-down attempt at description of the whole domain at a more abstract level
divorced from actual practice that might under- or over-determine the field.   

The Rhetoric of Interoperability: Urging Agreement.

The need for the development of common standards is promulgated by stressing the
importance of "interoperability."  A recent (March 2007) conference
<http://pdftohtml.markoer.org/pdf2html.php?url=http%3A%2F%2Fwww.comcare.org%2Fup
loads%2FInteroperability%20for%20Informed%20Emergency%20Response%20Final%20Repor
t.pdf> sponsored by COMCARE among others specifically focused on persuading
attendees of the importance of technological interoperability.  The keynote
address was given by the Under Secretary of Science and Technology Directorate,
Department of Homeland Security, in which he stated that his primary challenge
was the interoperability of first responders' systems in emergencies.  This call
was echoed by each of the other speakers in succession and it is also found in
other settings
<http://72.14.253.104/search?q=cache:n1UDmNniMokJ:web-services.gov/GSA12122005A.
doc+semantic+interoperability+emergency&hl=en&ct=clnk&cd=1&gl=us&client=firefox-
a> .  The solution to the interoperability problem envisaged by the agencies in
the field is the development of common standards and this could provide the
impetus for a full-blown ontology of disaster management.   

The rhetorical context of the need for common standards is the failure of
interoperability in the past.  This has become a pervasive history of recent
disasters, in the sense that a history is a set of reasons why things are
sub-optimal in a domain.  The reaction to this situation is the implementation
of a set of institutions to produce and manage consensus. 

Institutions Involved in Emergency Management Technologies: Manufacturing
Agreement.

The development of the standards outlined above is mediated through a set of
inter-related institutions that bring several perspectives to the process.  The
Department of Homeland Security plays host to the overarching model for
communication, the Emergency Interoperability Consortium brings together
industry players in emergency management, COMCARE represents practitioners in
the disaster field and OASIS manages the development and publication of
standards.

The Department of Homeland Security <http://www.dhs.gov/index.shtm>  sponsors
EDXL and co-hosts the National Information Exchange Model <http://niem.gov/>
(NIEM) with the Department of Justice.  NIEM aims to provide a framework for
information exchange through an online depository of XML-based languages and
namespaces that cover several domains, including emergency management.  The
elements and their attributes that constitute NIEM, accessible via a graphical
browser <http://niem.gtri.gatech.edu/iepd-ssgt/DataModelViewer.do> , are a
burgeoning taxonomy required for information exchange in settings relevant to
both the Department of Justice and the Department of Homeland Security, based on
the Global Justice XML Data Model (GJXDM).  NIEM is expressed as data components
that describe common concepts in the activities of the interested agencies.
NIEM specifies three types of component: Universal (such as person, address,
organization) that are used in all domains; Common that are exchanged between
many but not all agencies; and Domain-specific that are managed by a  "community
of interest".  Each domain -"a business enterprise broadly reflecting the
agencies, units of government, operational functions, services, and information
systems which are more or less organized or affiliated to meet common
objectives"- has a namespace and emergency management is one of these.  CAP and
EDXL are to be found in the domain.  The taxonomy building mechanism recognizes
that as new domains enter the model, particular components may become "Common"
and if all domains recognize a component it will then become "Universal".
Domain-specific namespaces inherit the component properties from Common and
Universal namespaces.  The Community of Interest is comprised of practitioners
and technical representatives who are authoritative representatives of their
domain and who collectively have a stake in NIEM information exchanges. 

The Emergency Interoperability Consortium <http://www.eic.org/>  (EIC) is a
principal member of the emergency management "community of interest" that groups
organizations and institutions that promote the creation of a national approach
for data interoperability and support the development of Web services and XML
technologies to effect that interoperability.  The members of the consortium are
heavily drawn from industry, though there are some not-for-profit organizations.

COMCARE <http://www.comcare.org/>  is also a principal member of the "community
of interest".  It is an organization of over a hundred members
<http://www.comcare.org/Members.html>  that are involved in emergency response
ranging from specialized medical associations to multinational corporations.
COMCARE supports the development of EDXL and hosts or co-hosts many
national-level meetings where issues regarding common standards in emergency
management are raised and EDXL is publicized and advocated.

OASIS acts as the body that coordinates the technical description of the
standards and publishes the element sets and schemata.  It has a technical
committee that is dedicated to Emergency Management (EM TC
<http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=emergency> ) that is
in turn composed of four subcommittees that oversee such things as the messaging
and notification standards and their adoption.  Proposed standards are presented
to the committee that examines them and publishes them for public comment.
After public comment has ended, the TC usually moves to ratification and then
publication.  

The propagation of common understanding through these organizations is supported
by their interpenetration: COMCARE and the Department of Homeland Security
(Disaster Management) are both members of EIC, many EIC members participate in
OASIS EM TC, and EIC is a member of COMCARE.  Thus the "community of interest"
is able to arrive at elements, attributes and relationships that support their
interventions in disaster management.  Together these institutions function as
intermediaries between the various interested parties in Emergency Management
creating a locus of homogeneity through committee-based procedures.  In this
sense, EDXL is an abstraction of the domain mediated via institutional
processes.  

It is possible to glimpse the institutional ontology work being done at NIMS
<http://www.fema.gov/emergency/nims/index.shtm>  by looking at their
documentation. The resource typing is done by NIC (and its expert bodies) and
then opened to public comment. They formulate their work like this: "The role of
the NIC is to establish interoperability of resources through consensus
definition for teams and equipment, and Knowledge, Skills, and Abilities for
individuals and team members." The degree of control they wish to exert is
expressed in their request that local communities not undertake their own
typing: in response to the FAQ (presumably coming from a local emergency
response community) "Does that mean we are supposed to do our own resource
typing, or what?" the NIC states that "No, you should not start 'typing' your
resources. Communities and jurisdictions should begin to use the resource typing
definitions to describe and inventory their resources using the Resource Typing
Definitions that are listed on the NIC website". It is difficult to find a
balance between central control and public involvement in ontology development.
On the one hand, an ontology is a standard and needs oversight. On the other
hand, it is supposed to represent the community's domain, meaning that the
community needs to express its universe in the ontology. In a demonstration of
how this process works, NIC relates how, after three years of development, the
resource list was put out to public comment. As a result of this public scrutiny
of their expert work the understandably recondite element "Fire Truck" was added
to their typed list of emergency resources!
<http://www.fema.gov/emergency/nims/faq/rm.shtm> 

The Need for Ontology:

The XML-based technologies that are the primary products of efforts in the US to
develop common standards for emergency management are meant to be
human-readable.  For example, the CAP transmits standardized understandings of
disasters to machines that render them on a screen for users to read.  This is
the case with GDACS, which gathers several CAP feeds (and other message formats)
in one place so that the humanitarian relief agencies can do one stop shopping.
The semantics of this type of system are supplied by the human being interfacing
with the application.  While these languages and protocols have developed
ontologies of the communications necessary to deal with emergencies, in the
sense that they are a theory about the constituents of that particular
domain-universe, they do not deliver on the promise of the Semantic Web.  In
this vision, ontologies would exist in a form that would allow inferences to be
made based on the description logic expressed in the RDF/RDFS and OWL stack.
Parts of EDXL have been written in OWL: Rex Brooks
<http://www.starbourne.com/About.html>  , a member of the EM TC, has developed
an OWL representation of the distribution element of EDXL (accessible at Cover
Pages as an Excel file <http://xml.coverpages.org/ni2005-05-19-a.html> ), that
contains the simple class relationships between the elements of this part of
EDXL.  However, it seems that this type of description is not required by NIEM
or OASIS, nor that this OWL file is going to interoperate with other OWL
ontologies.   

While the results of the work of this community of interest are chiefly a
hierarchical taxonomy of the elements needed in communications during the
Response Phase of emergencies, some movement is being made towards an
ontological representation of disaster management that would include type
information.   Brooks has initiated a move to what he calls a reference
information model
<http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2007_01_25#nidSH5>
(RIM) with OWL-DL representation that is supposed to be the way to maintain a
model driven architecture.  Brooks is suggesting that in order to coordinate the
family of EDXL vocabularies, a higher-level overall model is required.  A
posting to an ontolog forum
<http://ontolog.cim3.net/forum/ontolog-forum/2007-04/msg00191.html>  suggests
that the EM TC is developing the EDXL-RIM with an RDF Schema and an OWL-DL
specification.  On the face of it, this might conflict with the requirement
driven standards that have emerged to date as small pieces of the emergency
management jigsaw have been put into place.  Moreover, there is little
intimation that the force behind this ontological representation would be to
facilitate machine inference.

Meanwhile, other suggestions have been made for ontologies.  The ESW Wiki has an
active page on Disaster Management
<http://esw.w3.org/topic/DisasterManagement?highlight=%28management%29%7C%28disa
ster%29>  that is proposing the initiation of a W3C Incubator Group
<http://esw.w3.org/topic/Charter_for_a_Proposed_W3C_Incubator_Group>  to develop
an ontology.  The mailing list
<http://lists.w3.org/Archives/Public/public-disaster-management-ont/>
associated with this effort is currently working out areas of commonality and
difference, focusing initially on the scope of the domain.  There is no mention
of the language of representation of this ontology but its proposed connection
to W3C would suggest they might be looking at RDF/RDFS and OWL.  The expressed
goals are to develop a directory of ontologies, a common ontology and data
dictionary and best practices in the application of interoperability standards.
At this early stage the participants seem to be casting the net very wide to
encompass all attempts at disaster management vocabularies and the technological
expression of the ontology.

The Institute for Crisis, Disaster and Risk Management (ICDRM)
<http://www.gwu.edu/%7Eicdrm/>  at George Washington University has developed a
glossary
<http://www.gwu.edu/%7Eicdrm/publications/PDF/GLOSSARY%2002-19-2007.pdf>  of
terms for emergency management with the support of the Department of Veterans
Affairs.  As the name suggest, this is a flat alphabetical list, incorporating
some NIMS terms. At present their is no relational or hierarchical information,
though this glossary could provide input into an ontology of disaster management
if such a thing were attempted.     

In Britain, Essex County Council in conjunction with DIP
<http://kmi.open.ac.uk/projects/dip/>   prototyped Semantic Web Services that
include GIS Emergency Planning
<http://72.14.253.104/search?q=cache:Z_iCYYFc3hAJ:www.eswc2006.org/poster-papers
/FP42-Davies.pdf+developing+sws+for+e-government&hl=en&ct=clnk&cd=1&gl=us&client
=firefox-a> .  The prototype was modeled around a real event (a snow storm at an
airport) in order to provide real data in a decision support system.  The aim
was to assist the user in gathering information in a more timely manner and to
this end they developed an "emergency ontology" in Web Services Modeling
Language.  The "emergency ontology" was one of several that were incorporated
into the system.  A hierarchy of ontologies was developed, with an archetypal
ontology at the top that abstracted properties of resources to enable
interoperability.  For example, a "hospital" was abstracted to a "house"
archetype by describing it as a "container" for "people".  Automated "lifting
operations" translated the XML lower level ontologies into the archetypes in
OCML <http://citeseer.ist.psu.edu/motta98overview.html>  which were employed in
the application and then concepts were "lowered" into their native XML,
presumably to make use of the specific resources.  As they admit, this type of
architecture required a description of the native ontology, the archetype
ontology and the lifting and lowering operations: an intensive process.   On the
other hand the application seemed to work well enough to challenge for prizes in
the Semantic Web world.
<http://istresults.cordis.europa.eu/index.cfm?section=news&tpl=news&ID=88396> 

In Europe an RDF metadata ontology
<http://64.233.167.104/search?q=cache:UiXrGqhul7YJ:grid.ucy.ac.cy/reports/TR-04-
07.pdf+building+a+distributed+natural+disasters&hl=en&ct=clnk&cd=2&gl=us>  has
been developed to facilitate the dissemination of research about natural
disasters.  One of the goals of the EU-MEDIN <http://www.eu-medin.org/>  project
is the better management of disasters by making the results of research readily
available to those doing planning and response.  In order to make the library of
resources widely available, the metadata attached to the resources have been
automatically translated into RDF triples which thus become application
independent.  The resulting RDF schema contains a rudimentary ontology of
disaster types.
 
Other Approaches: 

Paola Di Maio has been forthright in suggesting that a schema is not an ontology
and that the current EDXL work is not at that level
<http://ontolog.cim3.net/forum/ontolog-forum/2007-06/msg00055.html> .  Di Maio
is advocating an "open ontology"
<http://72.14.253.104/search?q=cache:Wt-WV7gvDcEJ:opensource.mit.edu/papers/TOWA
RDS_AN_OPEN_ONTOLOGY_FOR_ER.pdf+di+maio+open+ontology&hl=en&ct=clnk&cd=1&gl=us&c
lient=firefox-a>  for emergency response given that there are different "views
of the world" among agencies that deal with emergencies.  The ontology she
envisions would have a "sufficiently generic common conceptual and semantic
framework" that could be developed using a lightweight methodology, the "open
ontology." Each domain expert would build their own sub-graph of the ontology
and would be able to review the work of others on their sub-graphs.  The
collaborative nature of the work would develop an ontology that reflected the
highest common denominators of different interest groups in the emergency
response field.  In effect this would be akin to a Wikipedia of the emergency
management ontology, though the tools used would be different.  It seems that Di
Maio believes that this type of collaborative process will produce an ontology
that is richer and more robust, as fully expressive of the entities, attributes
and properties in the field as possible.  She comments that building an ontology
is more than opening an editor and creating OWL or RDF files.  

Whether this type of approach produces better-engineered ontologies, when
compared to the committee and procedure based US model, is an open question
which will be decided empirically.  The goal of any semantic, machine-readable
ontology effort is to capture the universe it addresses in a way that allows
machine inference to take place and it would be instructive to see if this type
of approach approximates that universe more accurately.  It is unclear if an
open ontology method is that different from the formal approach since both are
attempting to gather the input of all interested parties.  Moreover, the open
source paradigm is, in reality, not so bazaar-like as often represented.  In
many software cases, the process has become hierarchical and institutionalized
mirroring traditional approaches to design.

At the same time, Di Maio is aware of the possible dangerous consequences of
this approach.  It may be that an open source method might not fulfill the
exacting requirements of a system that is to be deployed in life-saving
situations.  She suggests that traditional methods, that are risk averse and
highly structured, might have a natural advantage in this field.  To counter
this, she suggests frequent testing of the system
<http://ontolog.cim3.net/forum/ontolog-forum/2007-04/msg00198.html>  that will
validate the ontology that is produced.  

Interestingly, Di Maio addresses politics in her advocacy of an open source
approach.  She contends that the products of the OASIS process, CAP and EDXL,
are top-down with little public input and a preponderance of vendor influence.
She also mentions that one of the major players in that process, EIC, is
"US-centric."  Moreover, one of the requirements for an open ontology is that it
would be usable by alternative ontology languages, not just OWL.  This political
disagreement seems to be focused on the method of determining an ontology,
rather than its contents but the notion that an ontology is top-down imposition
strikes at the heart of the resources and properties that it comprises.  Since
the ontology is supposed to represent the domain, a version that does not
account for all participants will fall short of the ideal.  In effect, Di Maio
is saying that the domain is larger than the US approach acknowledges and that
any ontology coming from it will be the weaker for it.  In this regard it is
revealing that she draws attention to Sahana <http://www.sahana.lk/>  that is
not only an open source emergency management application, but also one developed
outside the US.  She points out that despite its success and longevity, it has
lacked a consistent data model or ontology that means that terms are used
inconsistently and interchangeably and that the developers are now acknowledging
the need for semantic integrity.

Sahana
<http://pdfdownload.04340.com/pdf2html.php?url=ftp%3A%2F%2Fftp.umiacs.umd.edu%2F
pub%2Flouiqa%2FPUB06%2FSahana6.pdf>  was developed as an open source software
application for disaster management in the aftermath of the 2004 Asian tsunami
and ascribes part of its success to the alignment of open source and
humanitarian principles.  The system has core modules that mimic in part the
scope of EDXL: an Organization Registry that tracks the agencies involved; a
Request Management System where request for aid are matched to offers of
support; a Shelter Registry that keeps track of the available shelters; and a
Missing Person Registry that acts like a bulletin board for both missing and
found people.  The system has been deployed several times and each time new
customizations are added to enrich functionality.  Functions that could be
implemented in the Semantic Web vision, such as matching requests and reports,
were effected manually in the 2006 Philippines deployment, though it would be
possible to automate the task within the Sahana system.  Sahana is effectively a
standalone application, with translators/scripts for imports of different data
types.  Di Maio's concern about this system is that it does not embody a uniform
ontology of disaster management, neither within the system, nor across the
disaster management domain.  The creators of Sahana comment that deployments of
the system sometimes do not account for the granularity of the data.  They give
the example of a crate of food: it might be known that it is arriving, but the
system might not support describing what the contents are.  This is a comment
about the ontology in the system and it is a shortcoming that Sahana seeks to
overcome by customizing the application during each deployment.  In effect, the
repeated deployment of Sahana and its iterations during these deployments are
developing an ontology for disaster management through use.  

Like Sahana, other systems have been developed or proposed on an ad hoc basis to
cope with emergencies.  After hurricane Katrina, a registry of missing persons
was developed (PeopleFinder) that employed a data model, or ontology, of missing
people expressed in XML, the People Finder Interchange Format
<http://zesty.ca/pfif/1.1/>  (PFIF ).  The element set contains some similarity
to that used in Sahana
<http://demo.sahana.lk/index.php?mod=mpr&act=addmp&type=missing>  , but there
are differences.  Whether these differences could be resolved or whether they
are expressions of the local nature of the disaster in which the systems were
used is unknown.

During the 2004 Asian tsunami a blog (SEA-EAT)
<http://www.firstmonday.org/issues/issue11_5/jones/>  was set up that developed
into an information exchange for missing persons, requests for help and news
updates.  This mechanism was simply human readable, a presentation of text that
could be scanned to gain information.  Soenke Ziesche has suggested that a wiki
could be developed with typed links and attribute pairs that would be amenable
to machine interrogation <http://www.xml.com/lpt/a/1683>  (the SEA-EAT creators
concede that a wiki would have been more appropriate for their purposes than a
blog, though they do not suggest machine readability.)  Ziesche contends that if
the links in the articles on the wiki were typed, it would be possible to query
the wiki to derive inferences - he gives an example of determining the number of
tents available in a district from three different articles that type the tents,
the districts and their super-districts.  Of course, in order to do this the
developers would need to define an ontology of the types that are relevant to
disaster management.  This remains the open question for this implementation.

Each of these community-based, collaborative systems contains elements of a
disaster management ontology.  Even David Stephenson's idea about using cell
phones in emergencies to generate situational awareness information (reported on
O'Reilly Radar <http://radar.oreilly.com/archives/2007/02/peer_to_peer_in.html>
) employ an ontology of the elements of disaster communication (the location
information in the technologies), though it does not expressly create one.  His
wider view
<http://stephensonstrategies.com/directory-of-major-blog-posts/10-21st-century-d
isaster-prep-tips-you-wont-get-from-government/>  of self-help in a disaster is
the most free-form expression of disaster management: ad hoc links and mash-ups
are created that exploit existing ontologies, without the need to develop a
top-down description of the domain beforehand.  In some ways this is similar to
the microformats <http://microformats.org/> movement that calls itself the
"lower case semantic web".  Discrete semantic elements are hacked into existing
systems to meet human need.  By their nature, these on-the-fly responses to
disasters are not institutionalized, but the recent ESW Wiki page
<http://esw.w3.org/topic/DisasterManagement?highlight=%28management%29%7C%28disa
ster%29>  and Di Maio's intervention in the OASIS
<http://ontolog.cim3.net/forum/ontolog-forum/2007-04/msg00198.html>  process
<http://ontolog.cim3.net/forum/ontolog-forum/2007-04/msg00198.html>  appear to
be the first step towards a higher level aggregation of these different types of
ontologies into a unified theory of disaster management.

Conclusion:

At the time of writing the field of disaster and emergency management is
experiencing a great deal of interest in the possibility of finding an
expressive ontology that represents the domain and the actors and agencies
within it.  The institutional activity in the US is at the level of application
ontologies and moving into the description of domain ontologies.
<http://64.233.167.104/search?q=cache:afxfbG_WpHEJ:www.deri.at/fileadmin/documen
ts/DERI-TR-2003-10-29.pdf+de+bruijn+enabling+knowledge+using+ontologies&hl=en&ct
=clnk&cd=1&gl=us&client=firefox-a>   In terms of their expressiveness, they are
akin to controlled vocabularies with some thesaurus-like relations made
explicit.  Because they are embedded in (governmental) institutional processes
they are being propagated to the emergency management community as important
standards that will overcome the problematic history of responses to disasters.
(There is a complex web of relations between these institutions. For example,
the Department of Homeland Security sponsors the emergent ontology of EDXL while
one of the department's constituents, FEMA, houses NIMS with its typed resource
list.  Some NIMS terminology, though not necessarily the typed resources, is
used in turn in the ICDRM glossary at George Washington University.)    At the
same time, these ontologies are the subject of some push-back, explicitly from
actors who see them either as lacking in expressiveness or representing US
concerns, and implicitly from the spontaneous generation of ontologies in
response to local emergencies that circumvent the barriers to adoption of the
Semantic Web <http://citeseer.ist.psu.edu/750854.html> . 

The Semantic Web vision of a system that could answer a complicated request at
the time of a disaster is far from realized.  For example, if an Emergency
Officer needed enough tents and food for 3400 people, deliverable in one day,
first by air to the local city, then by road to the disaster area accompanied by
fifteen distribution experts, the parts of this request would need at present to
be broken into separate items.  The required number of tents and amount of food
would have to be computed, the location of the items discovered, the logistics
put in place.  This would be done by interpreting human-readable messages
presented by the existing software applications, possibly glued together with
unifying scripts.  And perhaps this would be good enough: ad hoc integration
might satisfy the need.  Whether an overarching ontology allowing machine
inference is needed in this domain is still an open question, though it is being
pursued in some quarters.

It is also debatable whether any ontological effort will extend beyond the
Response Phase.  The construction of ontologies that determine disaster
mitigation and disaster recovery may be too politically threatening since they
would normatively prescribe our standards of living, what is desirable in the
planning of human communities and what is acceptable in their reconstruction.
The need for such ontologies is growing as the severity of the effects of
natural disasters is exacerbated by demographic and land use patterns that are
increasingly leaving communities vulnerable.
<http://www.gdrc.org/uem/disasters/1-what_is.html>   Perhaps this is why the
notion of resilience <http://www.ony.unu.edu/25Apr2006.html> , a parallel to
business continuity planning, is emerging as the leitmotif in disaster
management, the ability to absorb the inevitable disasters that will impact
human occupations and to recover as viable communities.   At present, however,
the field of emergency management is concentrating on developing a description
of the Response Phase, an ontology for coping with chaos as it were  It is
ironic that this is the area where islands of uniformity are to be found while
the description of ontologies of resilience that might mitigate disasters are
harder to come by. 

 


The information transmitted in this electronic communication is intended only
for the person or entity to whom it is addressed and may contain confidential
and/or privileged material. Any review, retransmission, dissemination or other
use of or taking of any action in reliance upon this information by persons or
entities other than the intended recipient is prohibited. If you received this
information in error, please contact the Compliance HelpLine at 800-856-1983 and
properly dispose of this information.
Received on Monday, 21 July 2008 14:11:50 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 21 July 2008 14:11:51 GMT