pelagios 2011

24
The Pelagios ontology workshop Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation Monika Solanki Department of Computer Science University of Leicester [email protected] http://http://commons.wikimedia.org/ Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Upload: monika-solanki

Post on 11-May-2015

700 views

Category:

Documents


0 download

DESCRIPTION

Presentation at the Pelagios Linked Data workshop March 24, 2011

TRANSCRIPT

Page 1: Pelagios 2011

The Pelagios ontology workshop

Semantic Reasoning over Qualitative Spatial Relationsfor Image Interpretation

Monika Solanki

Department of Computer ScienceUniversity of Leicester

[email protected]

http://http://commons.wikimedia.org/

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 2: Pelagios 2011

Talk outline The Pelagios ontology workshop

Outline

Motivation

Proposed framework

Conclusions and future work

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 3: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

Tracing Networks

investigates the network of contacts across and beyond theMediterranean region, between the late bronze age and the lateclassical period (c.1500-c.200 BCE) by interrogating materialobjects

seven archaeological case studies fully integrated with computerscience projects

programme sets technological networks in their greater social,economic and political contexts to expand our understanding ofwider cultural developments

these networks from the past can help us devise new and moreeffective ways of transmitting knowledge and information in ourdigital world

http://www.tracingnetworks.org/

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 4: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

Tracing Networks

Archaeologists study a wide range of material objects.

By tracking them at every stage of their production,distribution, use, and consumption across a largegeographical region, over a long time period, they cantrace the links between the people who made, used, andtaught others to make them.

The Chaîne opératoire

Cross-craft interaction

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 5: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

The TN-LOD cloud

Tracing Networks through Linked Open Data

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 6: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

Tracing Networks: Vocabularies

CIDOC-CRMIt provides definitions and a formal structure for describingthe implicit and explicit concepts and relationships used incultural heritage documentation.

An ontology of 86 classes and 137 properties for cultureand more.

International standard since 2006 - ISO 21127:2006.

The ontology has been encoded in OWL2.0, OWLDL andRDFS.

http://cidoc.ics.forth.gr/index.html

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 7: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

Tracing Networks: Vocabularies

CIDOC-CRMIt provides definitions and a formal structure for describingthe implicit and explicit concepts and relationships used incultural heritage documentation.

An ontology of 86 classes and 137 properties for cultureand more.

International standard since 2006 - ISO 21127:2006.

The ontology has been encoded in OWL2.0, OWLDL andRDFS.

Tracing Network vocabularies extend CIDOC-CRM

http://cidoc.ics.forth.gr/index.html

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 8: Pelagios 2011

Tracing Networks The Pelagios ontology workshop

Transformation Framework

ORM Reverse Engineering.

ECA Rule-based Transformation.

Ontology Instance Generation.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 9: Pelagios 2011

SEA: Semantic Explorer for Archaeology The Pelagios ontology workshop

SEA: Semantic Explorer for Archaeology

MotivationThe most time-consuming part of an archaeologicalinvestigation is the post-excavation analysis.

There is therefore a mileage in combining the task ofarchiving, querying and analysing the data within a singleframework.

Archaeological data is fragmentary. Inferencing capabilitiesof reasoners can be used to extract implicit knowledge andcontribute to their existing knowledge bases to completethe fragments.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 10: Pelagios 2011

SEA: Semantic Explorer for Archaeology The Pelagios ontology workshop

SEA: Semantic Explorer for Archaeology

A web application.

RESTful APIs for programmatically accessing the TN-LODcloud.

Interactive and global querying of linked datasets.

Data visualisations using user defined perspectives.

Statistical analysis using bespoke criteria provided byarchaeologists at runtime.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 11: Pelagios 2011

SEA: Semantic Explorer for Archaeology The Pelagios ontology workshop

Case study: Human representations

The scope of the project includes examining and analysinghuman representations on a range of object types and in arange of materials, such as bronze and pottery.

The project utilises details such as gestures and postures,dress and associated objects as keys to understandinghow identity and new understandings of society arecommunicated.

Raw data is collected through examining objects frompublished literature or in museum collections.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 12: Pelagios 2011

SEA: Semantic Explorer for Archaeology The Pelagios ontology workshop

Human representations: Informal queries

Example 1:

“Find images of riders who appear on the east side of objectsfound in Austria where the altitude of the excavation site is 500meters above sea level. I would also like to know the statisticaldistribution of the material and the technologies used for theproduction of these objects. I would like to visualise the resultsas a pie chart and see the distribution of the sites where theseobjects were found on Google Earth”.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 13: Pelagios 2011

SEA: Semantic Explorer for Archaeology The Pelagios ontology workshop

Human representations: Informal queries

Example 2:

“Find all objects which have images of individuals in the orantgesture who are wearing a triangular dress, earrings and whocarry a vessel on their head, where the vessel is supported bytheir left hand. I would also like to know the statisticaldistribution of the gender of these individuals according to thecountry in which the objects were found. I would like tovisualise the results as a tree map and see the distribution ofthe sites where these objects were found on Google Map”.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 14: Pelagios 2011

Motivation The Pelagios ontology workshop

Motivation/Problem statement

Capture “spatial” metadata about the content of images.

Spatial relation based on RCC and the cardinal/ordinaldirections.

Use ontologies to support annotation and consequentlysearch.

Multiple vocabularies to support the process.

Annotate only those regions of the image which arecontextually relevant and interesting.

Use “existing” spatial DL reasoners based on RCC-8.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 15: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Region Connection calculus: RCC-8

http://en.wikipedia.org/wiki/Region_connection_calculus

Defines a primitive reflexive and symmetric dyadic relationC(x; y), - region x connects with region y.

On top of this relation, a number of other dyadic relationsare defined.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 16: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Region Connection calculus: RCC-8

http://en.wikipedia.org/wiki/Region_connection_calculus

disconnected (DC)

externally connected (EC)

equal (EQ)

partially overlapping (PO)

tangential proper part (TPP)

tangential proper part inverse (TPPi)

non-tangential proper part (NTPP)

non-tangential proper part inverse (NTPPi)

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 17: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Cardinal and ordinal directions

http://en.wikipedia.org/wiki/File:Brosen_windrose.svg

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 18: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

A “coarse” notion of spatial fuzziness

X right_of Y: what does “right_of” entail in terms of cardinal andordinal directions?

north_west_of

south_west_of

west_north_west_of

west_of

west_south_west_of

right_of

south_east_of

north_east_of

east_south_east_of

east_of

east_north_east_of

left_of

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 19: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Vocabularies used

CIDOC CRM version 5.0.2 (2010) (crm)FOAF (foaf)Annotation vocabularies (ano)Dublic Core Metadata Initiative (dc)Pellet spatial RCC-8 vocabulary (spa)Cardinal-ordinal vocabulary (crd)Domain specific vocabularies to describe images.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 20: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Image representation: TBox

ImageInArt ⊑ foaf:Image, ⊑ crm:E38ImageInArt ⊑ ∀ hasImageRegion.ImageRegionImageRegion ≡ ImageInArt

⊓ ≥1hasInitEndCornerSelector⊓ ≥1hasAnnotation⊓ ∃hasAnnotation.ano:Note⊓ ∃hasInitEndCornerSelector.ano:InitEndCornerSelector

hasImageRegion ⊑ dc:hasPartisImageRegionOf ⊑ dc:isPartOfhasImageRegion ≡ isImageRegionOf−

hasInitEndCornerSelector ⊆ isInitEndCornerSelectorOf−

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 21: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Pellet Spatial

A Hybrid RCC-8 and RDF/OWL Reasoning and QueryEngine.

Provides consistency checking and query answering.

Supports all RCC-8 relations as well as standardRDF/OWL semantic relations.

Supports SPARQL querying over consistent sets of spatialrelations and non-spatial semantic RDF relations.

http://clarkparsia.com/pellet/spatial/

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 22: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Example

http://en.wikipedia.org/

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 23: Pelagios 2011

Proposed Framework The Pelagios ontology workshop

Conclusions

A preliminary approach has been presented to queryimage repositories based on the spatial orientation ofimage regions.

A coarse notion of fuzziness has been used in defining thespatial orientation of image regions relative to each other.

Cardinal/ordinal directional predicates along with RCC-8relations have been used.

Existing DL based spatial reasoners exploited.

Still WIP: lots more needs to be done.

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation

Page 24: Pelagios 2011

The Pelagios ontology workshop

Many Thanks!!!

Monika Solanki Semantic Reasoning over Qualitative Spatial Relations for Image Interpretation