Semantic enrichment of VGI using Linked_Data_Stanislav_Ronzhin_defence

Semantic enrichment of Volunteered Geographic Information using Linked

Data:a use case scenario for disaster

managementby Stanislav Ronzhin

Supervisor: Rob Lemmens

Professor: Menno-Jan Kraak

Reviewer: Marian de Vries

• Problem• Linked Data at work• Research Question & Objectives• Use case• Research steps• Results• Discussion• Conclusion

Content

Problem -1

“…information is very directly about saving lives…”

“we found the data was very rich and could be used for various other products that would aid in the response"

Sir John Holmes, The UN Emergency Relief Coordinator

“tweets were as accurate as official reports…; they were also at least two weeks faster”

Chunara et al., 2012

Mr. Andrej Verity, an UNOCHA veteran

Problem -2

From Minu Kumar LIMBU, 2012

Expert Survey: Knows vs. Uses

Problem -3

From Minu Kumar LIMBU, 2012

Why did the service not help?

Problem -4

“…lives were saved…”

Ushahidi Independent Evaluation

To sum up:

• Unstructured data

• Lack of interoperability

• Semantic heterogeneity

• Uncertainty & reliability

=Problems of

crowdsourced content

LD at work - 1

Linked Data to the rescue!

• Smart and elegant• Graph Data Model• Subject-Predicate-Object statements – LEGO!

LD at work -2Resource Description Framework (RDF)

Relational DB Graph DB

LD at work - 3

LD at work -5

Linked Data:• Structured data• Integration between any

sources• Formal semantics• Giant Open data

repository• Sophisticated

informational retrieval

LD at work -6

Problem:

• Unstructured data

• Lack of interoperability

• Semantic heterogeneity

• Uncertainty & reliability

• Keyword search

+

=• Can we cross the bridge with a 12 ton fire truck?• How to reach the closest operating hospital avoiding road

blockages and who is in charge at that place?

LD at work -7

Volunteered Geographic Information

To what extent can the Linked Open Data cloud help to semantically enrich Volunteered Geographic

Information in order to better answer queries in the context of crisis and disaster relief operations?

Research Question & Objectives

Objectives:

1. To integrate VGI into the LOD cloud

2. To evaluate methods for the construction of semantic queries

3. To evaluate the results

Use case-1

Chile earthquake, 2010

1. The 6th largest ever

2. Magnitude of 8.8

3. 80% of population

4. Tsunami

5. 525 casualties

Russia, 2010Haiti, 2010Chili, 2010

Christchurch, 2011Libya, 2011 Australia, 2011

Kenya , 2007South Africa, 2008 Louisiana, 2010

Serial INCIDENT TITLE INCIDENT DATE LOCATION DESCRIPTION CATEGORY LATITUDE LONGITUDE APPROVED VERIFIED

1636 200 need food and water in Laboule.

20/01/2010 10:46

Laboule, Port-Au-Prince

AIDE POUR LA FONDATION REGARS SISE A LABOULE COORDONER PAR PETIT HOMME STANEL NOUS AVONS ENVIRON 200 PERSONNES

~~~~~~~~~~~~~~~~~~~~~~~~~

We need help for the Regars Foundation located in Laboule, directed by Stanel Petit-Homme. We have about 200 people who need help.~~~~~~~~~~~~~~~~~Category: 4a. Health services

2b. Penurie d'eau | Water shortage, 2a. Penurie d'aliments | Food Shortage, 3c. Besoins en materiels et medicaments | Medical equipment and supply needs,

18.49282 -72.3069 YES NO

Use case-2

Research steps -1 Step 1: to convert Ushahidi into RDF

Research steps-2Step 1: to convert Ushahidi into RDF

Column name ValueSerial number 4349Incident title SERVICIO DE SALUD CONCEPCIÓN

FUNCIONANDO

Incident date 3/4/2010 11:08:00 PMLocation Concepcion, ChileDescription Hospital Guillermo Grant Benavente ,

Hospital Traumatológico ,Hospital de Lota ,Hospital de Coronel

Category 4a. Servicios de Salud, Latitude -36.8148Longitude -73.0293Approved YESVerified NO

Thematic

Spatial

Spatial

Temporal

Research steps-3 Step 1: to convert Ushahidi into RDF

The MOAC vocabulary

The Dublin Core vocabulary LinkedGeoData

Research steps-4Step 1: to convert Ushahidi into RDF

http://linkedgeodata.org/triplify/node988381631 http://linkedgeodata.org/triplify/way126614190 http://linkedgeodata.org/triplify/way223990111 http://linkedgeodata.org/triplify/way124859821

"http://observedchange.com/moac/ns/#HospitalOperating"

Thu Mar 04 23:08:00 CET 2010

Research steps -5Step 2: to construct queries for a semantic enrichment

Research steps -6Step 2: to construct queries for a semantic enrichment

LinkedGeoData – geometry and classification

Research steps -7Step 2: to construct queries for a semantic enrichment

DBpedia – background information

Research steps -8 Step 3: to evaluate emerged data management techniques

Before After

Ushahidi classKeyword search

MOACMachine readable

Multi criteria filtering

Literal dateKeyword search

DateTemporal distanceMachine readable

One point per reportKeyword search

Georeferencing Spatial querying

Interoperable geometry

Results-1Extraction of geometries of blocked roads

# TITLE DATE LOCATION DESCRIPTION CATEGORY

LAT/LONG

4730

Ruta 5 sur esta cerrada/ Ruta 5 South

Closed

3/11/2010 17:48

Rancagua, Chile

Road N. 5 closed due to the structural damaged

inflicted by the aftershocks

1a. Estructur

a Colapsad

a,

-34.034

2/-

70.7056

Results-2

Selection of officials of populated places where operating hospitals are located

Prioritizing of the reports based on the density of population.

Results-3

Conclusion

“Challenge is to provide an appropriate categorization (with sufficient explanation) so that volunteers can classify correctly from the beginning…”

An UNOCHA veteran

Conclusion

• MOAC – domain knowledge - WHAT

• LGD – location – WHERE

• Spatial data access

• LOD is an integrated dataspace

• Approach can be applied to any VGI

Discussion

• Integration with LOD mitigates human factor

• Need for LD interface

• Remote SPARQL endpoint is a black box

• GeoSPARQL implementation in Virtuoso

• GeoSPARQL lacks KNN

Thank you!Questions!?

Methodology – Software

• TripleGeo• SILK Workbench• RDF Online Validator• Parliament

• iSPARQL• Parliament

• Tabulator• SIMILE Widgets

Background – Linked Data resources• DBpedia – Wikipedia in RDF• GeoNames – Gazetteer, 10 million toponyms• LinkedGeoData (LGD) – OSM in RDF • WordNet & GeoWordNet - lexical database + Geo extension

http://lod-cloud.net/versions/2014-08-30/lod-cloud_colored.png

Methodology – data conversion to RDF

Methodology – data conversion to RDF

Background – Ontologies and Vocabularies

URI – to name things, conceptsOWL – to express formal semanticsSPARQL – query languageRDF – to wrap information

Hypothesis-4

Example of queries:• Find me all the KFC restaurants located in less than 1 km from

the school(s) where Barack Obama had classes.• Find me names of famous singers who lived in municipalities

along river Rhine and who used word “Rhine” in their songs.

LD in work -6

The MOAC vocabHaitian

Earthquake 2010 Chilean Earthquake 2010

Category number

English name

Category number Spanish English

translation

MOAC termsPrefix MOAC: <http://

observedchange.com/moac/ns/#>

Terms from other vocabularies

1 Emergency 1 Emergen

cia Emergency MOAC:Emergency

5a

Collapsed structure,

1a

Estructura Colapsada

Collapsed structure

MOAC:CollapsedStructure

1b Incendio Fire MOAC:Fire

1c Trapped people 1c

Personas atrapadas

Trapped people

MOAC:PeopleTrapped

1e Tsunami Tsunami

http://ontologi.es/WordNet/data/Tsunami

Report as a graph

Report as a graph

Results-2

Selection of bridges located in 10-km proximity to compromised bridges