What are the lessons for the technical community?

although hard-to-link formats are a problem, the regulatory setting of reusability is crucial: data needs an open license to begin to count as open data.

What is the main lesson for the technical community?

The perennial user interface issue is critical in this politicized context, as the ease with which ordinary citizens can access and query the data is a crucial factor for OGD’s value.

What is the challenge of integrating OGD with PSI?

The challenge is to manage the sudden influx ofheterogeneous data, often with minimal semantics and structure, tailored to highly specific task contexts.

What is the way to access data?

It uses predefined templates for translation into RDF, so that users must transform the spreadsheet into a template and then add metadata such as a dataset name, description, and URL.

What is the main issue in the debate about OGD?

Bottlenecks in Exporting OGD to the LDW Much discussion about transparency has focused on the unwillingness of public service providers to surrender control of their data.

What is the main purpose of the article?

OGD is not a rigid government IT specification, but it demands productive dialogue between data providers, users, and developers.

Who is the chief architect at Seme4?

Hugh Glaser is the chief architect at Seme4 and a visiting research fellow in electronics and computer science at the University of Southampton.

What is the way to link data?

Linking is also possible via linksliding; currently Geordi supports finding and asserting owl:sameAs, but future work will extend the range of properties that can be asserted.

(Open Access) Linked Open Government Data: Lessons from Data.gov.uk (2012) | Nigel Shadbolt

Published by the IEEE Computer Society

L I N K E D O P E N G O V E R N M E N T D A T A

Linked Open

Government Data:

Lessons from

Data.gov.uk

Nigel Shadbolt, Kieron O’Hara, Tim Berners-Lee, Nicholas Gibbins, Hugh Glaser,

Wendy Hall and m.c. schraefel, University of Southampton

A project to extract

value from open

government data

contributes to the

population of the

linked data Web

with high-value data

of good provenance.

to citizens’ needs. Even when governments

have exposed service provision to market

disciplines, they haven’t succeeded in pre-

senting data to citizens in innovative ways

to create new value streams. Privatized ser-

vice providers have preserved monopolies of

service design and provision.

As technology has increased the power

of data by facilitating linking and sharing,

and political thinkers have embraced trans-

parency and citizens’ right to data, this top-

down culture is being challenged. Many

governments now release large quantities

of data into the public domain, often free

of charge and without administrative over-

head. This allows citizen-centered service

delivery and design and improves account-

ability of public services, leading to better

public-service outcomes.

In the United Kingdom, transparency

is focused on Data.gov.uk, the public data

catalogue that points to thousands of data-

sets downloadable under a permissive open

government license. The datasets are often

in comma-separated value (CSV) format or

spreadsheets, but there is potential for in-

creasing their utility by linking them using

structured machine-processable formats.

Resource Description Framework (RDF)

is the format most integrated into current

thinking about future generations of the

Web, as its use of URIs allows data to be

identied by reference and linked with other

relevant data by subject, predicate, or ob-

ject. The use of Semantic Web standards

in open government data (OGD) was pio-

neered by Advanced Knowledge Technolo-

gies (AKT) in a precursor to the work de-

scribed here, and was reported to the UK

Parliament in 2007.

We refer to this vision as the linked-data

Web (LDW). The LDW is already well pop-

ulated through initiatives such as DBpedia,

the DBLP Computer Science Bibliogra-

phy, the London Gazette, the New York

Times, and the Comprehensive Knowledge

Archive Network (CKAN). The formalisms

and infrastructure are appearing according

ervices require data. In a top-down political culture where the state is

the service provider of rst resort, the state becomes a powerful data

monopoly, able to structure and homogenize the interactions between itself

and its citizens. Such one-sided interactions are expensive and unresponsive

IS-27-03-Ohar.indd 2 5/18/12 4:32 PM

MAY/JUNE 2012 www.computer.org/intelligent 3

to linked data principles set out by

Tim Berners-Lee some time ago (www.

w3.org /designissues/ linkeddata.

html), but vital research issues still

need to be addressed.

First, we need to understand how

to build or reuse ontologies easily and

appropriately for particular applica-

tions. Second, we want query meth-

ods that scale across the unbounded

Web, not just within small islands of

well-structured data. Third, we need

visualization and browsing tools, and

fourth, we need to populate the LDW

to increase the network effects of large-

scale linking. These objectives drive

the fundamental research of the

EnAKTing project (www.enakting.

org), funded by the UK’s Engineer-

ing and Physical Sciences Research

Council (EPSRC).

OGD will make an important con-

tribution to the LDW. Its quantity

will help deliver the network effects

expected from the LDW, its prov-

enance is clearer than that of many

other types of data, and it is often

seen as high quality, trustworthy, and

neutral.

Representing OGD in RDF and

linking to other datasets presents im-

portant research challenges, including

• discovering appropriate datasets for

applications,

• integrating OGD into the LDW,

• understanding the best join points

for diverse datasets—that is, the

points of reference the databases

share, which are extremely valu-

able for linking—and

• building client applications to con-

sume the data, including interfac-

ing with real-world users.

In this article, we use the EnAKTing

approach to develop an integrated

account of how to bring OGD into

the LDW. EnAKTing’s focus is the

LDW as a whole, but here we focus

on the population of the LDW with

OGD from Data.gov.uk, looking in

turn at these four issues.

Discovering and

Migrating Data

The adoption of OGD for use in the

LDW will depend on its availability,

and a necessary first step into ex-

panding the LDW with OGD is the

data discovery process. There have

been a number of services supporting

the location of public sector informa-

tion (PSI), including Data.gov in the

US and Data.gov.uk in the UK. Tools

research and development is per-

mitting the translation of PSI data-

sets into RDF and the generation of

RDFa (RDF with attributes) catalogs,

while the UK government is expos-

ing linked-data endpoints of available

PSI for reuse.

However, innovative uses of PSI

transcend borders; meteorology or

transport applications, for example,

need data from more than one na-

tion. The LDW will be an important

mechanism for data convergence, as

shown by the Open Knowledge Foun-

dation’s CKAN—a registry of open

data available for public use with a

common cataloging schema built on

a few metadata terms—and by the

European statistical service Eurostat,

which has amalgamated thousands

of datasets with their metadata for

download from its website. However,

there is no single facility for retriev-

ing related resources from the portals

of the various nations, or for search-

ing intelligently across regional, na-

tional, and supranational sources.

EnAKTing has proposed an archi-

tecture—not yet fully implemented—

for integrating PSI catalogs via the ac-

tivities and components essential for

discovery. Architectures of this type

allow the presentation of catalogs in a

standardized form, facilitating search

and retrieval across resources.

The rst phase of this architecture

involves downloading and transform-

ing catalogs with retrievable records

into a common schema language for-

mat, whereas the second addresses

semantic heterogeneity with schema

matching and statistical analyses of

ontology structures. Once common

ontologies are in place, the search en-

gine layer can be developed, allowing

distributed querying and federated

search and retrieval.

Initial work has tested this archi-

tecture, using approximately 7,000

records taken from Data.gov.uk, the

US site Data.gov, and the Australian

national PSI catalog. Records were

converted from native format into

RDF, each detailing some 14 to 25

metadata elds, and stored in an RDF

triple-store.

The initial translation was inten-

tionally minimal, reecting the cata-

logs’ original contents and preserving

the underlying arrangement of data.

This reveals the need for data nor-

malization. For example, temporal

data such as release or modication

dates were not always represented

with a universal standard. With thou-

sands of ambiguous dates, classiers

need to be applied to the data before

evaluation and comparison of the re-

sources referred to in the catalogs are

possible.

Integrating OGD into

the Web of Linked Data

Once datasets have been discovered,

they must still be integrated into the

linked data cloud. An application we

developed for EnAKTing provides an

example of such integration, bring-

ing together six government datasets,

covering the work of individual MPs

(members of Parliament) and of Par-

liament as a whole, crime, mortality

and health statistics, and geographi-

cal data from the Ordnance Survey.

This application lets users investigate

IS-27-03-Ohar.indd 3 5/18/12 4:32 PM

4 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMS

L I N K E D O P E N G O V E R N M E N T D A T A

a particular geographical region.

Only the Ordnance Survey material

was in RDF.

Publication and Consumption

of the Datasets

Using well-known ontologies such

as Dublin Core, Friend of a Friend

(FOAF), and the Statistical Core Vo-

cabulary (Scovo) eased the modeling

overhead. Scripts were written to con-

vert data from spreadsheets into RDF,

and the Jena Semantic Web Frame-

work was used to convert the HTML

and XML , making data linkable

without determining the semantics.

For instance, data commonly con-

tains terms that make perfect sense to

experts in the eld but are opaque to

the rest of us; the health datasets used

in our application included the codes

SHA Code and Org Code, which can

only be understood by someone au

fait with UK National Health Service

(NHS) administration. Such prob-

lems multiply across datasets, requir-

ing an ontological alignment stage.

In our application, this involved the

correct identication of

owl:sameAs

relations across a dimension linking

the datasets. Administrative geogra-

phy provided the link, via MPs’ con-

stituencies, NHS trusts, and so on.

The alignment can be complex. For

instance, to align the health statistics,

we needed to use the Google Maps

API to get the coordinates of NHS

administrative units, and then query

the Ordnance Survey data manually

using string matching for the corre-

sponding Parliamentary constituen-

cies. The time dimension adds further

complication to administrative geog-

raphy. Parliamentary constituencies

are regularly redrawn in response to

demographic change, and different

data sources deal with this in differ-

ent ways; the Ordnance Survey ad-

ministrative geography stores only

the latest classication.

However, when issues such as

changes of semantics do not occur, our

techniques allow incremental on-the-

fly updating for data consumption.

Many of the applications discussed

in this article visualize a single store,

using data harvested and processed

into RDF by EnAKTing research-

ers. This data, along with the as-

sociated visualization based on the

current contents of the store, can be

refreshed at any point. Other applica-

tions query Data.gov.uk in real time.

So, for example, See UK (http://apps.

seme4.com/see-uk/) imports the UK’s

monthly crime data into its store, and

the view is always of the latest gures

(see Figure 1).

The Value of Place

for Linking Datasets

Geography provides an intuitive way

to align datasets—no surprise, as

governments generate PSI about the

territory over which they have juris-

diction, so the data has an implicitly

geographical dimension. The LDW is

well-stocked with geographical data;

the Geonames service manages eight

million URIs for geographical re-

sources. Therefore, where there is an

authoritative geographical knowledge

base available, as in the UK, geog-

raphy is an irresistible join point for

datasets.

(See the sidebar, “Related

Work on the Linked Data Web.”)

In our application for EnAKTing,

the region gives context for the dis-

played data and is the central point

from which we link to the LDW. New

views of the data or concepts generate

new searches and presentations on the

basis of aggregations that make sense

in the new contexts; for instance, hav-

ing moved up the geographical hierar-

chy from a constituency to a county,

the application can present the statis-

tics (such as crimes committed) rele-

vant to the county as a whole.

This approach will not work with

some types of territory, such as Parlia-

mentary constituencies, which don’t

map easily onto the administrative

geography of the UK. Yet, if we can

Figure 1. See UK, showing relative crime gures for a ward in Southampton. The

pie chart shows comparisons between it and neighboring wards normalized by

population, and the user can select gures and comparisons for particular classes

of crime from the drop-down menu.

IS-27-03-Ohar.indd 4 5/18/12 4:32 PM

MAY/JUNE 2012 www.computer.org/intelligent 5

establish that one entity is completely

contained within another (for ex-

ample, the Parliamentary constitu-

ency of Fareham within the county of

Hampshire), we can discover relevant

data and present it to the user, ideally

exploiting existing LDW resources or

bringing more geographical resources

onto the LDW (see Figure 2).

To help with this kind of reasoning,

EnAKTing has developed a service

(http://geoservice.psi.enakting.org) to

support the discovery of geographical

resources pertaining to the UK on the

LDW by querying containment rela-

tions.

This service exploits knowl-

edge about instance equivalence

that is already available via corefer-

ence systems such as SameAs (http://

sameas.org). It normalizes the data,

translating the

os:contains relation

into two statements, a

h as-part and

part-of, to produce a structure

such as that in Figure 2, allowing the

service to infer containment via re-

sources from difference datasets us-

ing

owl:sameAs.

Reasoning Services

Geolinking services are only one kind

of reasoning needed to enrich linked

data. As another example, EnAKTing

has developed a backlinking service

pen government data (OGD) is becoming increas-

ingly important across the globe, although currently

most initiatives involve making data in proprietary

formats downloadable. Surveys have shown that there are

relatively few attempts to combine OGD with the linked-

data web (LDW),

and that Data.gov.uk and Data.gov are

unusual in their commitment to the LDW vision. Many

other important and interesting developments have been

more opportunistic, including the creative use of Open

Street Map data in the aftermath of the Haiti earthquake

in 2010, while initiatives such as the Open Government

Partnership (www.opengovpartnership.org) have begun

to spread best practices even further.

The work closest to our project aims to migrate Data.gov

to the LDW. The Tetherless World Constellation (TWC)

Linked OGD portal

also recognizes data-publishing stages

for OGD on the LDW:

• the catalog stage, where an inventory of datasets is

created;

• the retrieval stage, where a snapshot of the dataset’s

online data file at a point is input to a Linked OGD

converter; and

• the conversion stage, where the data is converted to RDF

in a layered manner that allows many of the conversion

issues and bottlenecks to be sidestepped.

An initial automatic conversion is done by the portal, and

enhancements such as mapping ad hoc database column

names to common properties can be done by users. Many

of its linking strategies were anticipated by the Data-gov

wiki.

The strategy of the LOGD portal has been to fos-

ter an LOGD community by actively engaging users with

demos.

The Data-gov wiki is a social Semantic Web platform that

has produced more than 5 billion triples, covering topics

such as government spending, environmental records, and

statistics on the cost and usage of public services. It goes

through a series of steps similar to those just outlined,

including conversion of data into RDF, enhancement and

linking by declaratively associating URIs in related contexts

(done both automatically and by hand), and designing ap-

plications and demos to address the important issue of data

consumption. The Data-gov wiki limited its efforts to well-

formed comma-separated value (CSV) les, and so was able

to sidestep several conversion issues. It also took a lightweight

approach, with a minimal and extensible conversion to

preserve the structure and content of the raw data and no

more. The TWC team did not use properties from existing

ontologies, to avoid manual moderation, but properties

used in converted RDF data were dereferenceable (that is,

accessible from their URIs) to terms in well-known ontolo-

gies (such as Friend of a Friend and Dublin Core) or RDF and

XML pages generated by Semantic MediaWiki. The Data-

gov wiki also focused on provenance, and was able to use

this as a join point, linking by derivation- and version-based

provenance associations.

Evangelos Kalampokis and his colleagues have also ex-

ploited the social Web, using OGD to enrich data mined

from social networking and microblogging sites—for exam-

ple, linking tweets from high-crime areas in the UK to the

crime data from http://police.uk for those regions.

The aim

of this work is to allow policy makers to assess public opin-

ion and predict public reaction. Kalampokis’s team’s linked

data architecture integrates OGD with data mined from

the social Web, to enable the collection of OGD related

to a specic set of criteria that the decision-maker provides.

The integrator integrates and stores as RDF the social data

with objective data related to the specied target group,

as well as the variables related to social data and real-world

objective facts coming from government data. The improve-

ment of the OGD comes via augmentation from the social

Web, rather than from the integration processes used by

EnAKTing and TWC LOGD. Crowdsourcing (obtaining data

from a distributed group of citizens) is clearly an important

way forward.

References

1. E. Kalampokis, E. Tambouris, and K. Tarabanis, “A Classica-

tion Scheme for Open Government Data: Towards Linking

Decentralized Data,” Int’l J. Web Engineering and Technology,

vol. 6, no. 3, Inderscience, 2011, pp. 266–285.

2. L. Ding et al., “TWC LOGD: A Portal for Linked Open Government

Data Ecosystems,: J. Web Semantics, vol. 9, no. 3, Elsevier,

2011, pp. 325–333.

3. L. Ding et al., “Data-gov Wiki: Towards Linking Government

Data,” Proc. AAAI Spring Symp. Linked Data Meets Articial

Intelligence, AAAI, 2010, pp. 38–43.

4. E. Kalampokis, M. Hausenblas, and K. Tarabanis, “Combining

Social and Government Open Data for Participatory Decision-

Making,” Proc. 3rd Int’l Conf. eParticipation (ePart 11), LNCS6847,

Springer, 2011, pp. 36–47.

Linked Open Government Data: Lessons from Data.gov.uk

Citations

Acceptance and use predictors of open data technologies: Drawing upon the unified theory of acceptance and use of technology

New Horizons for a Data-Driven Economy: A Roadmap for Usage and Exploitation of Big Data in Europe

The dual effects of the Internet of Things (IoT) : A systematic review of the benefits and risks of IoT adoption by organizations

State-of-the-art in open data research: Insights from existing literature and a research agenda

Driving innovation through big open linked data (BOLD): Exploring antecedents using interpretive structural modelling

References

TWC LOGD: A portal for linked open government data ecosystems

A classification scheme for open government data: towards linking decentralised data

Data-gov Wiki: Towards Linking Government Data.

Unlocking the potential of public sector information with semantic web technology

Combining social and government open data for participatory decision-making

Related Papers (5)

Benefits, Adoption Barriers and Myths of Open Data and Open Government

Linked Data - the story so far

Linked Data: Evolving the Web into a Global Data Space

Open Government Data

Open data: An International comparison of strategies

Frequently Asked Questions (9)

Q1. What contributions have the authors mentioned in the paper "Linked open government data: lessons from data.gov.uk" ?

Q2. What are the lessons for the technical community?

Q3. What is the main lesson for the technical community?

Q4. What is the challenge of integrating OGD with PSI?

Q5. What is the way to access data?

Q6. What is the main issue in the debate about OGD?

Q7. What is the main purpose of the article?

Q8. Who is the chief architect at Seme4?

Q9. What is the way to link data?