ontology virtualization for smart data -- a semantics perspective on open data sharing

AMisunderstanding Semantic Web Value Proposition Towards Ontology Virtualization

Ontology Virtualization for Smart Data

A Semantics Perspective on Open Data Sharing

Krzysztof JanowiczSTKO Lab, University of California, Santa Barbara, USA

RDA Metadata and Semantics Workshop, Feb. 23, 2015

1/18


HowWe Think About Data

Big Data As The New Natural Resource

http://www.ibmbigdatahub.com/infographic/big-data-new-natural-resource

2/18


HowWe Think About Data

Big Data As The New Natural Resource

http://www.ibmbigdatahub.com/infographic/big-data-new-natural-resource

A suitable analogy?Natural, i.e., not man-madeExhaustible, finite quantityRenewable, replenishableConsumed, alteredBuilding block

If we don’t even understand what data are, how should we make sense out of them?

2/18


Retrieval

The Data Retrieval Problem Is Real

Even the major data hubs such as Data.gov still rely on keyword-based searchand have unreliable, incomplete, and missing metadata. For this type ofretrieval problems, even ’a little semantics goes a long way’ (Hendler 1997).

3/18


Sensemaking

Sensemaking is Difficult – Fitness for Puspose is Key

There is no shortage of data, butfinding data that is fit for a certainpurpose is difficult.Data as statements not as truth.Heterogeneity is caused by culturaldifferences, progress in science,viewpoints, granularity, ...semantics does not come for free.Lack of provenance informationSensemaking requires morepowerful semantic technologies andontologies (compared to IR).

4/18


Interoperability

Meaningful Analysis and Synthesis is Difficult

Ensuring that data is analyzed andcombined in a meaningful way is farfrom trivial.What if the information on how touse the data would come togetherwith these data?Focus on smart data instead of(merely on) smart applications.The purpose of ontologies is not toagree on the meaning of terms but tomake the data provider’s intendedmeaning explicit.

A little experiment: The statement all rivers flow into other water bodiesis not useful because it is ’true’1, but because...?

1 It is not; rivers can flow into the ground or just dry up entirely before reaching another water body.5/18


The Semantic Cube

The Semantic Cube

http://goo.gl/fBHie6

6/18


The Semantic Value Proposition

Value Proposition

Why use Semantic Web, Linked Data, and Ontologies?

Federated queries over multiple data sourcesUnique global identifiers easy conflation and deduplicationTransparent data model; reduces the need for guessingNo data silos, no API restrictionsMany pre-defined lightweight vocabularies (ontologies)Smart data reduces the need for smart applicationsMachine reasoning supportNo need for agreement, ontologies make hidden assumptions explicitDoes away with the data – metadata distinction!

7/18


The Semantic Value Proposition

The Smart Data Argument

One of the key arguments underlying the Linked Data paradigmis to make data smart, not applications. Instead of developingincreasingly complex software, the so-called business logicshould be moved to the (meta)data. The rationale is that smartdata will make all future applications more usable, flexible, androbust, while smarter applications fail to improve data alongthe same dimensions.

(http://goo.gl/fBHie6)

8/18


Semantic Search and Metadata Enrichment

Smart Data enables Semantic Search

Creating enriched, semantically-lifted Linked Data on top of Esri’s ArcGIS Online

9/18


Linked Data Exploration

Flexible Pattern-based Data Exploration

An user-friendly interface on top of the DBpedia SPARQL endpoints.

10/18


Linked Data Exploration

Flexible Pattern-based Data Exploration

An user-friendly interface on top of the ADL Gazetteer SPARQL endpoints.

11/18


Ontology Design Patterns

Ontology Design Pattern in a Nutshell

Modular but self-containedbuilding blocksSome patterns are strategiesReusable and extendibleEven huge ontologies can bemodularized using ODP (forexample DOLCE)No need to import full ontology andall ontological commitmentsDifferent types of patterns, e.g.content vs. logicalHow many patterns are there?

12/18



A (More Complex) Semantic Trajectory Pattern

A pattern for discrete trajectories of people, wildlife, vessels, and so forth.

13/18



Ontology Design Patterns Can Be Specialized

Trajectories that model the research cruises of scientific vessels14/18


Micro-Ontologies

AMicro-Ontology for Cruises

Combining the InformationObject, Event, Vessel, and Trajectory patterns

15/18


Micro-Ontologies

W3C Semantic Sensor Network XG Ontology

16/18


Micro-Ontologies

The Ontology Standartization Argument

Given the early success of data format standardization, weassume that standardizing meaning (via ontologies) is lessdifficult and more persistent than aligning and translating localmicro- ontologies. What if standardization is the more difficulttask?

(http://goo.gl/2e751)

17/18


Ontology Virtualization

Towards Ontology Virtualization

In analogy to hardware virtualization: given a set of ontology designpatterns and their combination into micro-ontologies, we can abstract fromthe underlying axiomatization by:

Dynamically reconfiguring patterns in a plug&play styleBridging between different patterns an micro-theoriesProviding ontological views and semantic shortcuts that suitparticular provider, user, and use case needs by highlighting or hidingcertain aspects of the underlying ontological modelMap between major modeling styles,e.g., the use of instancesversus classes

How do we handle different ontological commitments?Quine: To be is to be the value of a (bound) variableExample: Transportation is moving goods from one location to another.

18/18

ontology virtualization for smart data -- a semantics perspective on open data sharing

Science

databig data

linked data

shortage of data

butfinding data

guessingno data silos

major data hubs

ontology virtualizationhowwe

little semantics