SemanticallyAware SpatioTemporal Data Analysis for Humanitarian and Natural Crisis Management

1
Semantically-Aware Spatio-Temporal Data Analysis
for Humanitarian and Natural Crisis Management

• Dr Kristin Stock
• Centre for Geospatial Science
• University of Nottingham

2
Emergencies

• The River Trent is likely to flood who should
  be evacuated?
• Normal food crops in Liberia have failed where
  do I need to deliver emergency food supplies?
• There has been a chicken pox outbreak in
  Edinburgh - where are the high concentrations of
  children?

3
Introduction The Problem

• All of these scenarios can be aided by geographic
  information.
• Requires time, expertise, significant manual
  effort.
• Need to identify and search for
• data and
• operations that must be performed to produce the
  end result.

4
Introduction The Vision

• The user would express requirements in natural
  language.
• The requirements would be automatically
  interpreted.
• Data and operations to meet requirements would be
  automatically identified and combined.
• ? Result returned to user.

5
Introduction The Solution

• Builds on and refines existing architectures.
• The Semantic Web
• Ontologies that describe data and their
  semantics.
• Semantic Web Services (self-describing software
  modules).
• Registries that describe resources.
• Allows web services to be automatically
  discovered, interpreted, combined and invoked.
• But the full vision has not yet been realised

6
Outline

• Three main research areas
• Expressing and transforming user objectives into
  formal, semantically equivalent expressions.
• Decomposing user objectives and identifying
  semantically equivalent web services.
• Implementing an ontological registry to support
  semantic matching.
• Simple flood evacuation example.

7
The Research Problems

• How to express the real world problem?
• How to automatically map the real world problem
  to a chain of spatio-temporal processes?
• How to express the semantics of a spatio-temporal
  process?
• How to automatically map a spatio-temporal
  process to a chain of spatio-temporal operations
  (or other processes)?
• How to express the semantics of a spatio-temporal
  operation?

Real World Problem
3
1
2
Spatio-Temporal Analysis Process a
Spatio-Temporal Analysis Process b
Spatio-Temporal Analysis Process c
Spatio-temporal processes may also be composed of
other spatio-temporal processes
1 loop until
2
Spatio-Temporal Operation x
Spatio-Temporal Operation y
Spatio-Temporal Operation z
8
Expressing User Objectives (1)

• The River Trent is likely to flood who should
  be evacuated?
• Using current approaches, in order to solve this
  problem, the user would need to know
• The structure of available data.
• The function of available operations.
• How the operations might contribute to achieve
  the goal.

9
Expressing User Objectives (2)

• Would be better if the user could express the
  objective in natural language.
• Use Natural Language Processing.
• Data requirements map against available domain
  ontologies.
• Operational requirements transform into
  OWL-S/SWRL expressions.

10
Transforming Objectives into OWL-S/SWRL
Expressions (1)

• OWL-S is a language for expressing web service
  semantics.
• Semantics are expressed by specifying inputs,
  outputs, preconditions and results.
• SWRL is a logic language allows conditions to
  be expressed within OWL-S.

11
Transforming Objectives into OWL-S/SWRL
Expressions (2)

• Map to ontologies of
• Goals (evacuate, alert, find) maps to required
  outcome.
• Situations (flood, tsunami, famine, infectious
  disease) maps to model of affected area.
• Use OWL-S/SWRL expressions stored in ontologies
  to derive semantics of user objective.

12
Identifying Implemented Web Services to Achieve
Objectives (1)

• Available Web Services have different levels of
  granularity.
• Example high level objective FindPeopleToEvacuat
  e.
• Web services at this level are not likely to be
  available, so the objective must be decomposed
  and semantically compared with web services.

13
User Objective
Interpret and Formalise
OWL-S/SWRL Process Expression of Objective
Decompose
If not semantically equivalent, continue
decomposition
Determine Semantic Equivalence
OWL-S/SWRL Process Expression of Sub-objective
OWL-S/SWRL Process Expression of Web Service
The Decomposition and Semantic Matching Process
14
A Decomposition of the FindPeopleToEvacuate User
Objective
15
A Decomposition of the FindVulnerablePeople User
Objective
16
Identifying Implemented Web Services to Achieve
Objectives (2)

• The semantic matching process may prompt the user
  for further input.
• Community Services web service may ask which
  services to evacuate
• Hospitals?
• Aged Care Homes?
• Libraries?

17
Semantic Matching between Services and
Sub-Objectives

• Requires ongoing research
• Logical equivalence rules.
• Classification of spatial operations.
• Logical specification of spatial functions.

18
Ontological Registries (1)

• Will implement an ontological registry to
  demonstrate and test the research.
• Current Spatial Data Infrastructures (SDI) use
  registries that are semantically limited.
• An ontological registry will describe the
  semantics of all the resources in the SDI.

19
Registry Resources

• Data
• Web Services
• Ontologies
• Domain Ontologies
• Goal Ontologies
• Situation Ontologies
• All other information for SDI operation.

20
Ontological Registries (2)

• Ontological registries store and derive the
  relationships between registry resources.
• Will assist in the process of decomposition and
  semantic matching between web services with
  different levels of granularity.
• Example if a web service has semantics that are
  too general, follow the links to child web
  services.

21
Summary Dynamic Spatio-Temporal Analysis for
Emergency Management

• The user expresses an objective in natural
  language.
• The objective is mapped to ontologies and
  transformed into formal expressions.
• The formal expression is semantically matched
  with services and progressively decomposed.
• Outcome relevant services are executed and a
  result provided to the user

22
KnowledgeScope

• A knowledge infrastructure for eScience.
• A computational framework for
• expressing
• managing
• discovering
• annotating and
• utilising scientific knowledge.

23
KnowledgeScope Knowledge Modelling

• Scientists create knowledge (publications,
  theories, models, data) and tag it to describe
  it.
• Scientists use a wiki-based interface to assist
  with tagging.
• Scientists can create their own tags within the
  tagging upper-ontology.

24
KnowledgeScope Knowledge Infrastructure

• The system generates and updates networks that
  connect the resources together.
• Results in a dynamic, evolving representation of
  the scientific knowledge in KnowledgeScope

25
KnowledgeScope Visualisation

• Scientists can view knowledge in KnowledgeScope
  on the basis of space (maps), time (timelines) or
  themes (concept maps).
• Knowledge in any of the views can be filtered
  according to the other views.

26
KnowledgeScope Querying and Discovery

• Scientists will be able to express goals and
  context in natural language.
• The system will interpret the request and provide
  analysis and visualisation that is tailored to
  requirements.
• The system will also identify other research that
  may be of interest based on the use of similar
  techniques or patterns in other areas.

27
KnowledgeScope Enhancements

• Will add richer handling of research theories,
  clusters of scientists, paradigms, schools or
  thought and other collaborative aspects to
  scientific knowledge.

28
Outcomes

• Collaboration with semantics researchers to
  develop KnowledgeScope (grant application).
• Development of ideas on semantically-aware
  dynamic spatio-temporal analysis this work will
  continue...

29
Thank You!!