Boris Krstović - News 3.0 : Online Publishing And Semantic Web
E G 3.0 THROUGH SEMANTIC WEB NATURAL LANGUAGE … · Like Government 2.0 that is associated with...
Transcript of E G 3.0 THROUGH SEMANTIC WEB NATURAL LANGUAGE … · Like Government 2.0 that is associated with...
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 64
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
ENABLING GOV 3.0 THROUGH SEMANTIC WEB, NATURAL
LANGUAGE PROCESSING AND TEXT ANALYTICS
Islam A. Hassan, E-Government Group, Insight Centre for Data Analytics, National
University of Ireland, Galway, Republic of Ireland
Adegboyega Ojo, E-Government Group, Insight Centre for Data Analytics, National
University of Ireland, Galway, Republic of Ireland
Abstract
The notion of “Government 3.0” (Gov3.0) is gradually emerging among policymakers as a
label for next generation ICT-enabled technology innovation in government and successor to
“Government 2.0” (social media or web 2.0 based) initiatives. The Gov3.0 phenomenon has
been largely associated with all forms desirable attributes for future government institutions
such as increased agility and innovation capacity. However, only few scholarly works have
attempted to provide conceptualizations or analysis of this emerging phenomenon. This paper
offers such conceptualization, by describing Government 3.0 as Semantic Web (SWEB) or
Web 3.0 enabled innovation in government. While there are existing SWEB-based applications
in the government domain, the diffusion of this family of applications in the government has
been slow if not stagnated. We argue that the reason for this problem is largely due to the lack
of tools and domain specific resources for automatic semantic annotation of existing
government-related resources and contents on the web and social web. To address this
challenge, we describe how text-analytics and Natural Language Processing tools could be
used for information extraction and annotation of existing government resources on the web.
In addition, we describe how the semantic resources generated from our web-scale automated
annotation approach could be used to build two exemplar Gov 3.0 applications. Finally we
discuss the challenges in developing this Gov3.0 infrastructure.
Keywords: Government 3.0, Ontology based, information extraction (IE), Natural Language
Processing (NLP), Public Services, t-Government.
Introduction 1
With the adoption of the “Government 3.0” agenda by the Korean Government with the vision
to create a “transparent, competent and service-oriented” government [Nam 2013], there is a
growing interest in both the policy and academic arena to deconstruct what this concept may
stand for in general. In the Korean framework, Gov3.0 is characterised by two high-level
goals: 1) providing customized services tailored to various needs and demand, and 2) creating
new jobs and re-boosting development. While the Korean Government is arguably the only
country that has openly declared a “Government 3.0” agenda so far, practitioners in general
have associated this label with the next stage of e-government evolution (i.e. after Government
2.0 and Web 2.0). Like Government 2.0 that is associated with Web 2.0-based innovation,
Web 3.0 or the Semantic Web is expected to play a major role in enabling Gov3.0 [Nam
2013].
The emergence of the Semantic Web concept in 2001 marked an important stage in the Web’s
evolution. As stated in (Tim Berners-Lee et al. 2001), Semantic Web it is "an extension of the
current Web in which information is given well-defined meaning, better enabling computers
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 65
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
and people to work in cooperation." This proposition has not yet been fully realized, and
although many efforts have been made in this direction, much remains to be done. However,
over the years, a number of semantic-web applications have emerged in the government space.
For instance in (Medjahed et al. 2003), an approach for managing the e-government services
using web services and the ontologies to make e-government services machine interpretable
and processable was described. In (Klischewski 2003), two possible areas for using the SWEB
in the e-government domain were presented including the markup of informational resources,
in which all the informational resources to be made accessible through e-government
applications are annotated using ontologies and interrelated to resources. A semantic approach
enabling the publishing of statistical data gathered from different government agencies was
presented in (Hoxha & Brahaj 2011) which included providing a semantic representation of
the data based on an ontology, and then serialising the data as RDF triples following Linked
Data principles. The work described methods for querying single or combined dataset and
how to visualise the results. Similar work was reported in (Gheorghiu & Nicolescu 2011)
where semantic web based mash-up application providing access to governmental data was
offered as an interactive tool able to query certain SPARQL endpoints of interest using
semantic web standards in order to visualise, analyse, and enhance the results.
Despite the above examples, the diffusion of this family of applications in the government has
been slow if not stagnated. In our opinion, the reason for this problem is largely due to the
lack of tools and domain specific resources for automatic semantic annotation of existing
government-related resources and contents on the web and social web. Specifically, to
progress towards the Semantic Web, there is a need to convert existing and new web contents
to forms that can be understood by both humans and also understandable by machines. The
semantic annotation (or markup) of Web documents is the first step towards adapting Web
content to the Semantic Web. Providing the information elements that currently make up the
Web with a well-defined meaning would, inter-alia improve its contextual search capabilities,
increase interoperability between systems in ‘collaborative’ contexts and, when combined
with Web services, ultimately enable near automatic composition of applications (Sheth et al.
2002) (Tse-Ming Tsai et al. 2003). Unfortunately, most Web contents remain unstructured (i.e.
textual information) making this process difficult and cost of annotation high. Addressing this
problem requires the availability of Natural Language and Text analytics tools to
automatically extract domain-relevant information from textual data and creating domain-
specific annotation resources based on the extracted information.
There are few past efforts such as (Misuraca et al. 2012) (Lin et al. 2009) that aimed to
address this challenge by developing domain ontologies used along with the Information
extraction and Natural Language Processing (NLP) tools to identify structured information
from the raw text. Specifically, in (Alfonseca & Manandhar 2002), General Architecture for
Text Engineering (GATE) components and NLP tools were used to extract crime information
from police reports, newspaper articles, and victims’ and witnesses’ crime narratives
automatically with the output presented as a meaningful summary for police investigators.
These enabled the police investigators’ quick comprehension of crime incidents without
having to read an entire report. In (Asahara & Matsumoto 2003), structural patterns expressed
in terms of regular expressions combined with lexical conditions were used to detect the
typologies of provisions contained in a normative document and extracted its related
arguments.
A plausible explanation for the paucity of NLP applications in the government domain is the
high manual engineering efforts required in developing domain-specific rules needed to
meaningfully extract information of interest from text sources associated with traditional NLP
methods. With the emergence of domain independent or open information extraction (IE)
techniques, interest on how to exploit these tools as light-weight domain-specific IE tools are
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 66
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
growing. This paper describes one such effort aiming leverage a well-known “generic or
open” entity and information extraction tools to identify relevant terms and relations
associated with an e-government domain vocabulary related to public services.
Our proposed approach consists in using open information extraction (OIE) in combination
with a generic named entity recognition (NER) tools to extract public service information
from the different governmental websites. First, we extended the DBpedia Spotlight as a
generic NER tool to support extraction our domain-specific terms. Second, we used domain-
specific ontology to map the relations on the extracted triples from the OIE tool. Third, the
extracted entities and relations were used in populating Public Service Ontology.
The paper is organized as follows: Section 2 summarizes the background and the different
techniques for Information Extraction. Our ontology-driven Entity and Information Extraction
approach is presented in Section 3. Section 4 presents two applications scenarios or use cases
for the proposed approach. Section 5 presents the challenges associated with the
implementation of our approach. Concluding remarks are presented in Section 6.
Background 2
Semantic Web 2.1
The desire to extend the capabilities of the Web for publishing of structured data is not new,
and can be traced back to the earliest proposal for the World Wide Web (Tim Berners-Lee
1989) and subsequent papers on the topic (Berners-Lee 1992). Trends foreseen at the early
stages of the Web's existence included “Evolution of objects from being principally human-
readable documents to contain more machine-oriented semantic information” (Berners-Lee
1992), which can be seen as the seed of an idea that became known as the Semantic Web. The
vision of a Semantic Web has been interpreted in many different ways (Tim Berners-Lee et al.
2001; Marshall & Shipman 2003). However, despite this diversity in interpretation, the
original goal of building a global Web of machine-readable data remains constant across the
original literature on the subject. According to (Berners-Lee 1999) “The first step is putting
data on the Web in a form that machines can naturally understand, or converting it to that
form. This creates what I call a Semantic Web – a web of data that can be processed directly
or indirectly by machines”.
The Semantic Web provides a common framework that allows data to be shared and reused
across applications, enterprises, and community boundaries. Its well-defined data semantics
enable computer agents and humans to work in cooperation (Tim Berners-Lee et al. 2001).
Recent efforts in the World Wide Web Consortium (W3C) to implement Semantic Web
(Gruber 1995) have spurred interest in the use of ontologies for information modelling and
knowledge representation. Ontologies provide shared domain models that are understandable
to both humans and machines. They describe a set of concepts and relationships between
them. Ontologies provide a controlled vocabulary of terms that can collectively provide an
abstract view of the domain (Schreiber & Swick 2006; Uschold & Gruninger 2009). Such a
shared understanding of the domain greatly facilitates querying of data and increases recall
and precision. Semantic Web technologies and ontologies are being used to address data
discovery, data interoperability, knowledge sharing and collaboration problems. Software
agents can then be used to construct and provide dynamic services on the web.
Ontologies can be described in RDF (Resource Description Framework) [16, 17] which
provides a flexible graph based model, used to describe and relate resources. An RDF
document is an unordered collection of statements; each with a subject, predicate and object
(triples). These statements describe properties of resources. Each resource and property can be
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 67
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
identified by a unique URI (Uniform Resource Identifier), which allows metadata about the
resource to be merged from several sources. RDF has a formal specification and it is widely
being used in a number of standards. It provides a common framework for expressing
information, so it can be exchanged between applications without loss of meaning. RDF
Schema (RDFS) adds taxonomies for classes and properties. It allows expressing classes and
their relationships (subclass), and defining properties and associating them with classes. It
facilitates inferencing on the data based on the hierarchical relationships.
Fig.1 The Linking Open Data cloud diagram
OWL (Web Ontology Language) [(Hayes & Ian n.d.; McGuinness & Harmelen n.d.) provides
extensive vocabulary along with formal semantics and facilitates machine interpretability.
OWL is much more expressive than RDF or RDFS, allowing us to build more knowledge into
the ontology. For example, cardinality constraints can be imposed on the properties of an
OWL class. OWL is designed as a specific language to define and describe classes and
properties within ontology. It has many predefined built-in functionality. For example, an
ontology can import other ontologies, committing to all of their classes, properties and
constraints. There are properties for asserting or denying the equivalence of individuals and
classes, providing a way to relate information expressed in one ontology to another. These
features, along with many others, are important for supporting ontology reuse, mapping and
interoperability. Also, OWL is a standard and is supported by the standard organization W3C.
We have used the Web Ontology Language OWL to define the ontologies and RDF is used to
describe the actual instances within the knowledge base. As the semantic web grows, more
and more ontologies will be available in OWL. There will be a wide variety of development
tools available for integrating the different OWL ontologies and doing intelligent reasoning.
OWL provides 3 sub 5 languages with increasing levels of expressiveness and complexity.
They are OWL Lite, OWL DL (Description Logics) and OWL Full. We have used mostly
OWL Lite due to the reason that it is less complex than DL and Full. Also it suffices for our
purpose. More tool support is currently available for OWL Lite than others. A comprehensive
list of OWL Lite language constructs is available at (McGuinness & Harmelen n.d.).
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 68
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
By publishing Linked Data, numerous individuals and groups have contributed to the building
of a Web of Data, which can lower the barrier to reuse, integration and application of data
from multiple, distributed and heterogeneous sources. Over time, with Linked Data as a
foundation, some of the more sophisticated proposals associated with the Semantic Web
vision, such as intelligent agents, may become a reality (see Figure 1).
Information Extraction 2.2
To be able to manipulate vast amounts of unstructured data effectively, automated systems
require efficient and accurate methods to derive information structures directly from text. The
purpose of adding structure to otherwise flat text is to generate a partial representation of
content in a form that can be effectively manipulated by the computer (Small & Medsker
2013). Specifically, most applications typically require representation that captures key events
reported and the attributes of these events, including their role in analysing a corpus.
Information extraction is the field that primarily deals with text.
Information Extraction addresses a variety of problems including: identifying relations from
textual content (Embley et al. 1998; Buitelaar & Ramaka 2005) and automatic instantiation of
ontologies and building knowledge bases tools (Alani et al. 2003). In domain specific context,
information extraction could be used for obtaining information from specialized literature such
as biomedical literature (Kamegai 2002).
Traditional methods on IE have focused on the use of supervised learning techniques such as
hidden Markov models (Freitag & McCallum 1999; Skounakis et al. 2003), self-supervised
methods (Etzioni et al. 2005) and rule learning (Soderland 1999). These techniques learn a
language model or a set of rules from a set of hand-tagged training documents and then apply
the model or rules to new texts. Models learned in this manner are effective on documents
similar to the set of training documents, but extract quite poorly when applied to documents
with a different genre or style. This process was expected to be simpler than manually creating
patterns and rules by hand and therefore faster and more accurate than the pattern matching
technique. BBN’s statistical language model used this approach for their MUC system which
performed extremely well (Miller et al. 1998) at MUC and is currently at the core of their
leading IE system (Identifinder). However, the effort to annotate training data for each new
domain was found to be more challenging than expected (Small & Medsker 2013).
The semi-supervised learning (SSL) methods use smaller sets of annotated data than what is
used in supervised learning (SL) methods, and they are augmented with large amounts of un-
annotated data. The typical semi-supervised learning (SSL) process is that annotations of some
examples, named entities, events, and relations can be used to find more examples and thus
more patterns from unannotated text. Unsupervised learning (UL) methods attempt to glean
information automatically from the texts themselves, also called Open Information Extraction
(Dalvi et al. 2012; Etzioni et al. 2008). As the name suggests, this approach to IE does not
require a pre-specified vocabulary or ontology (Fader et al. 2011), nor does it necessarily need
training data or rules. Usually, this approach facilitates domain independent extraction of
assertions. This paradigm is often considered to be liberal in a sense that essentially any texts
between two entities’ mentions are considered as a relation. Obviously, this implicitly
promotes the recall while accepting a level of noise in the extraction results (Lin et al. 2009).
Named entity recognition and disambiguation 2.3
The Named Entity (NE) recognition and disambiguation problem has been addressed in
different research fields and they agreed on the definition of a named entity, which is an
information unit described by the name of a person or an organization, a location, a brand, a
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 69
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
product, a numeric expression including time, date, money and percent found in a sentence
(Grishman 1996). The Supervised Learning (SL) techniques used a large dataset manually
labelled. In the SL field, a human being usually trains positive and negative examples so that
the algorithm computes classification patterns. SL techniques exploit Hidden Markov Models
(HMM) (Bikel et al. 1997), Decision Trees (Borthwick & Sterling 1998), Maximum Entropy
Models (Hoffart et al. 2011), Support Vector Machines (SVM) (Asahara & Matsumoto 2003)
and Conditional Random Fields (CRF) (McCallum & Li 2003). The common goal of these
approaches is to recognize relevant key-phrases and to classify them in a fixed taxonomy. The
challenges with SL approaches are the unavailability of such labelled resources and the
prohibitive cost of creating examples. Semi-Supervised Learning (SSL) and Unsupervised
Learning (UL) approaches attempt to solve this problem by either providing a small initial set
of labelled data to train and seed the system (Ji & Grishman 2006), or by resolving the
extraction problem as a clustering one. Other unsupervised methods may rely on lexical
resources (e.g. WordNet), lexical patterns and statistics computed on large annotated corpus
(Alfonseca & Manandhar 2002).
The NER task is strongly dependent on the knowledge base used to train the NE extraction
algorithm, for example by leveraging on the use of DBpedia (Bizer et al. 2009), Freebase3 and
YAGO (Suchanek et al. 2007)ontologies. In addition to detect a NE and its type, efforts have
been spent to develop methods for disambiguating information unit with a URI.
Disambiguation is one of the key challenges in this scenario and its foundation stands on the
fact that terms taken in isolation are naturally ambiguous. These methods generally try to find
in the surrounding text some clues for contextualizing the ambiguous term and refine its
intended meaning. Therefore, a NE extraction workflow consists in analysing some input
content for detecting named entities, assigning them a type weighted by a confidence score
and by providing a list of URIs for disambiguation. Initially, the Web mining community has
harnessed Wikipedia as the linking hub where entities were mapped (Hoffart et al. 2011;
Kulkarni et al. 2009). A natural evolution of this approach, mainly driven by the Semantic
Web community, consists in disambiguating named entities with data from the LOD cloud. In
(Nadeau & Sekine 2007), the authors proposed an approach to avoid named entity ambiguity
using the DBpedia dataset.
Automatic Annotation 2.4
Annotation is the process of adding additional information to any other information such as
information in a book, document, online record, video etc. In linguistics, annotation is the
process of adding additional linguistics information such as morphological, syntactic,
semantic etc. to available linguistics forms to make these forms more descriptive, Annotation
of textual document is to identify the concept with the help of domain ontologies (domain
because semantic or concept analysis is mostly domain specific i.e. one thing defined for one
domain may have difference sense or concept in another). For this purpose, a robust text
analysis technique is required which will be composed of identification of object in a text,
identification of relationship between these objects and analysis on how these objects and their
relationships combine to form a concept. A lot of work has been done in the field of semantic
annotation using ontology as main guide, but there is still no complete automatic semantic
annotation tool available that has a good accuracy due to inherent ambiguity in Natural
Languages. There are many tools available for semantic annotation of textual document like
GATE (Cunningham et al. 2001), KIM (Alani et al. 2003), Melita (Ciravegna et al. 2002),
MnM (Vargas-Vera & Motta 2002), Magpie (Domingue & Dzbor 2004) etc. but none of the
tools are totally automatic. Furthermore, these systems perform annotation on words and
terminologies to identify real world objects and their relationship in the text. None of them
provide annotation above word level (Ahmed 2009).
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 70
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
Approach 3
As shown in Section 2, most of the current NER and IE tools provide only limited, generic
capabilities for information extraction and do not support domain specific content processing.
We describe here how we approach the problem, (see Figure 2). First we present the abstract
Framework Architecture and with focus on the semantic annotation and enrichment
component. In our work, we propose to use the Open Information Extraction (OIE) tools to
extract generic triples (statements) out of the raw text. Second, these triples will be processed
by DBpedia Spotlight, which will serve as a named entity recognition (NER) tool to extract
and annotate the public services documents or corpus (see figure [3]). Finally we populate an
ontology of public services, by mapping both the entity and the relations to the ontology to
produce the RDF that will be used as domain specific semantic resources to be harnessed in
our two Gov 3.0 applications examples.
Fig.2 The Framework architecture.
The main components of the proposed solution architecture include:
The Domain specific language resource – We started with building the domain specific
language resource, this language resource will be used in extending the NER tool to recognize
the public service terms. We built a seed corpus and worked on annotating it and semi-
automatically transforming it from unstructured to structured form.
Domain Specific Ontology - This ontology is needed for building the graph by mapping the
correct relations to the extracted entities. We have used a conceptual model for the public
service, based on the public service ontology, being developed under the auspices of the ISA –
the Interoperability Solutions for European Public Administrations programme Figure [4].
Named Entity Recognition (NER) - As described in the background section, the NER tool is
the component that is responsible for recognizing and disambiguating the extracted entity by
the open information extraction tuples, we tried to use DBpedia spotlight for doing this task.
DBpedia Spotlight is a tool for automatically annotating mentions of DBpedia resources in
text. Thus, it provides a generic domain independent NER tool that needs to be extended by
domain specific language resource in order to recognize the public service terms, there are two
possible ways to extend the DBpedia spotlight described below:
Using Wikipedia Infoboxes and DBpedia extraction engine.
Using N-Triples files for building indexes and Spotters training.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 71
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
Open Information Extraction (OIE) - The open information extraction is mainly used for
extracting the tuples automatically out of the raw text. It is a generic and domain independent.
We explored three of the famous state of the art OIE tools ReVerb, OLLIE, and ClausIE
during our literature review.
Tuples – This constitute the major outputs from the extracted information produced by the
OIE tool. The tuples are represented in the traditional Subject, Predicate Object format (S, P,
O) format.
Entities, Relations mapping and ontology population - This component is responsible for
populating the ontology (public service ontology in our case) with the recognized entities by
the NER, mapping the tuples into domain-specific relations based on the ontology, and
producing the Extracted information in RDF/XML format.
Fig.3 The Information Extraction System architecture.
The process for applying the framework is as follows:
1. Governmental websites and document are used as input for the OIE tool.
2. The OIE tool automatically extracts the information from the text in a form of tuples.
3. Transform tuples into domain specific entities and relations:
a. Use the extended NER to recognize and disambiguate entities out of the
tuples.
b. Map the tuples independent predicates into domain-specific relations based on
the ontology (i.e. the public service ontology).
c. Populate the ontology with the entities (3.a) and the (3.b).
The output of this process is the domain-specific extracted information in RDF/XML format.
Use case – Automatic Annotation of Public Service Descriptions 4
The goal in this section is to present two concrete applications of our solution. Before
presenting these applications, we discuss in more details how the domain-specific language
resource for public services domain annotation was created.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, London, UB8 3PH
Islam A. Hassan and Adegboyega Ojo 72
Enabling Gov 3.0 through Semantic Web, Natural Language Processing and Text Analytics
We used the Core Public Service Ontology for documenting services in Europe as the basis for
creating our language. Using this language resource used for the information extraction will
help in populating the public service ontology using the information extracted from other
governmental public service documents like PDF documents or web documents.
Fig.4 UML diagram for the Core Public Service Vocabulary. All classes and properties are in the CPSV namespace
unless otherwise indicated.
Extracting information and representing them in a linked structured data form will open up
opportunities for useful applications that can leverage these information, for instance
improving citizens experience in exploration of public service information on the web.
Specifically, Citizens can enjoy new applications, which are built on the populated ontology to
get better access for the public services. At the same time, governments could use it for
carrying out sentiment analysis or public opinions analysis on different public services
mentioned on the social media or web. These kinds of applications enhance the delivery of the
public services and enable better understanding the citizen’s needs. As illustration, we show in
Figure 5 the extraction of public service information from a governmental website. The figure
shows important attributes annotated in the text like the public service name, the different
inputs and we can use the page URL as the channel of delivering this service. In Appendix B
we can see the RDF/XML that produced by the purposed framework as an output of the
information extraction processes and after populating the result on the public service
vocabulary “ontology”.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 73
Title of the Paper
Next we describe the two applications involving: 1) the intelligent navigation and exploration
of public services on government portals and 2) monitoring of public services mentions on
social media, described in details in Sections 4.1 and 4.2.
Fig.5 Information Extraction of a public service information from it’s website.
Public Service intelligent navigator 4.1
As indicated above, the semantic markup of Web documents is the first step towards adapting
Web content to the Semantic Web. The embedded semantics approaches like Microformats
and RDFa create a link between the Semantic Web and (social) Web containing the
governmental pages (in our scenario). This link helps in consuming the semantic data from
third party applications such as crawlers, or search engines. However, until now the human-
readable presentation of these semantic meta-information has been set aside. From the
citizen’s point of view, the actual interpretation of a rendered document relies on:
1. The contextual knowledge provided by the hyperlinks contained in a document; 2. The background knowledge of the citizen that helps him to bind unlinked information
to known concepts. Citizens need to use the document’s internal and external
semantics to lead them to a better understanding of the public service documents
content. For making it possible, the document's representation must be augmented
with making explicit this external knowledge in a human readable way.
The public service intelligent navigator is conceived as a plugin for the Internet browsers that
enables citizens to consume the Public Service Linked Data generated from automatic
annotation of public service pages. To demonstrate the use of this tool, consider a citizen
searching for information about the driving test service in Ireland. After locating the
government webpage describing the sought service, the public service intelligent navigator
plugin in the browser will provide notification telling the user other complementary
information about the service like driving test service (see Figure 6). Clicking over the plugin
icon will popup three tabs containing information related to this service like, the related
services, required for and notes.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 74
Title of the Paper
Fig.8 The public service intelligent browser.
In the first tab the user will find information about the related services as in Figure [6] services
like “Apply for Deriver Theory Test”, “Find a Driving Instructor”, etc.
Under the “required for” tab the user will find all the services that require the output of the
current service for as an input to get it. In the last tab “notes” users will find the some advice
regarding taking the service that could be extracted from the social media as citizens now
share their experience of using the public services frequently “sharing knowledge”, you also
might find notes from the original website about the service under this tab too. Note that these
tabs are generated based on the relations described in the public service ontology presented in
Figure 4.
Information about any particular service could be retrieved from a triple data store or by using
embedded semantics, with the information published in the website. In the first case, we crawl
the governmental public service websites then use the extracted obtained information in
populating the ontology and stored in as linked data in the data store. There are also many
methods for embedding semantics as mentioned above. In the case of embedded semantic the
public service intelligent browser will extract the embedded semantic content from a page by
checking the document's RDFa statements and citizens will be able to request more
information about these statements by selecting any of them. When doing so, inferences from
inside and outside the page are executed for that statement, a contextualized view of the
obtained knowledge is presented and the related information within the document (statements
with the same predicate) displayed too (see Appendix A).
Monitoring Social media 4.2
Social media produce high velocity stream of information on the web. Twitter for example
produces millions of messages per day and hundreds tweets per second. The simplicity of
using it motivates people around the world to produce textual contents at a rapid rate. This
large volume calls for an automated interpretation to flexibly and quickly respond to shifts in
sentiment or rising trends. This way one can almost directly respond on phenomena arising in
society than traditionally and that will open for the citizens “unimagined opportunities to do
more for themselves (Johnston & Hansen 2011).
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 75
Title of the Paper
Fig.9 Twits that mentioned a Public Service
Driven by rising citizen expectations and the need for government innovation, social media
has become “a central component of e-government in a very short period of time” (Bertot et al.
2012). As social media exists almost in the entire world, the sentiment expressed by users of
social media is written in different languages. The social media connect the entire world and
thus people can much more easily influence each other and that could help in the
interoperability.
The social media analytics using text analytics such as Information Extraction, named entity
recognition tools to extract the public services information, combined with sentiment analysis
tools allow us to determine citizens feedback about the different services (see figure 9). For
example, tweets mentioning the “driving test public service” in Ireland are shown in the
figure. The output of the social media analysis could be used again to help the other citizens in
getting the service and also will help the government in understanding the citizen needs.
Mentions or entities of interests could be automatically identified in tweets or from any social
media sources such as Facebook pages, using the approach described in Section 3 with the
extracted information property semantically annotated based on some ontology or vocabulary
and stored in a knowledge base like an RDF store.
Challenges 5
We summarize here the two categories of challenges we confronted in the implementing this
solution. The first relates to the difficulties in developing the domain-specific language
resource and the second is linked to the choice of the Open Entity Extraction tool – the popular
DBpedia Spotlight.
Language Resource Challenges: 5.1
Manual effort in creating the domain specific language resource - Creating a domain specific
language resource for extending the named entity recognition still takes manual effort, albeit
less than the effort spent in the other approaches where traditional information extraction
systems is used. In our case we searched for the proper corpus that could be utilized as a
valuable seed for the language resource, while working on the data transformation and
mapping the documents to the proper ontology with the help of a domain expert was
challenging.
Semantic Challenges - Identifying synonyms names, sentence relations, across different
agencies in the same government, across government levels, across boarders; In the case of the
cross-border synonyms, there is a need to deal with the multi-lingual issues.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 76
Title of the Paper
DBpedia Spotlight Challenges: 5.2
Data Quality - DBpedia is highly dependent on the of cross-ontology mappings. This affects
the data quality of the extracted information and consequently the performance of DBpedia
spotlight. The quality of the extracted data is increased by providing quality assurance tools
for the Wikipedia community and to notify the authors about contradictions between the
content of infoboxes contained in the different language versions of the article. There are also
efforts on interlinking DBpedia with other knowledge bases such as Cyc and YAGO to apply
semi-automatic consistency checks for Wikipedia content. Crowdsourcing quality assessment
is also used over DBpedia to solve this challenge (Medjahed et al. 2003).
Resources Demand - Creating a spotlight model locally requires setting up Apache Hadoop
and Apache Pig, for hadoop-based indexing process. This requires an expensive computational
resource to be able to handle the English Wikipedia dump. These resources are not available in
most government IT ecosystem currently.
Poor Documentation - The third challenge is about the completeness of the documentation.
The documentation does not cover the different components of the DBpedia spotlight that.
Consequently, time is wasted in figuring out a solution for most of the problems faced while
working with DBpedia spotlight. However, the provided mailing list is a valuable source for
seeking solutions.
Conclusion 6
The work conceptualizes the notion of Government 3.0 as a Web.30-enabled innovation in the
government arena. Although the usage of Gov3.0 transcends technological issues (like in the
Korea Government case), we have sought to focus on the technical aspect of Gov3.0 to
highlight the enabling role that NLP and text analytics plays in the generation of semantic
contents on the semantic web. In fact, we attribute the apparent stagnation in semantic web
adoption in government and arguably other domains, to the lack of NLP tools that could
automatically generate semantic annotations guided by relevant ontologies. However, we
have also shown through our approach that successful application of NLPs tools is contingent
on the availability of domain specific language resources and domain adaptation (or specific)
frameworks. We have also shown that an effective Gov 3.0 application should enable
intelligent interactions between citizens and government online resources. In addition to
addressing the challenges outlined in Section5, our future work will focus on developing
government-specific language resources and relevant vocabularies to enable the processing
and annotation of different types of government information on the web. Furthermore, while
we have employed DBpedia spotlight as our NER tool, our recent experiments show that, a
detailed comparative analysis of domain-specific NER approaches may be beneficial to
improve both the accuracy and recall of the extracted information. A major implication of this
is work is the availability of a concrete language resource for government public services
domain that could be freely exploited by other researchers to build Gov 3.0 applications or
solutions. Another major side effect of the work is the availability of templates for describing
government services in Wikipedia. Consequently, the open community could define and co-
create public services descriptions on Wikipedia. This will significantly impact on the quality
of annotations obtained through DBpedia Spotlight.
References 7
Ahmed, Z., 2009. Domain Specific Information Extraction for Semantic Annotation. Available
at: http://is.cuni.cz/studium/eng/dipl_st/index.php?doo=detail&did=74742.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 77
Title of the Paper
Alani, H. et al., 2003. Web based Knowledge Extraction and Consolidation for Automatic
Ontology Instantiation. Available at: http://eprints.soton.ac.uk/258325/1/Alani-
SEMANNOT-camera-ready.pdf.
Alfonseca, E. & Manandhar, S., 2002. An unsupervised method for general named entity
recognition and automated concept discovery. … Conference on General …. Available
at: http://www-users.cs.york.ac.uk/~suresh/papers/AUMFGNERAACD.pdf.
Asahara, M. & Matsumoto, Y., 2003. Japanese Named Entity extraction with redundant
morphological analysis. In Proceedings of the 2003 Conference of the North American
Chapter of the Association for Computational Linguistics on Human Language
Technology - NAACL ’03. Morristown, NJ, USA: Association for Computational
Linguistics, pp. 8–15. Available at: http://dl.acm.org/citation.cfm?id=1073445.1073447.
Berners-Lee, T., 1999. Weaving the Web: The Past, Present and Future of the World Wide
Web by Its Inventor (with M. Fischetti). Available at:
http://scholar.google.com/scholar?q=Weaving+the+Web%3A+The+Past%2C+Present+a
nd+Future+of+the+World+Wide+Web+by+its+Inventor&btnG=&hl=en&as_sdt=0%2C
5#0.
Berners-Lee, T.J., 1992. The world-wide web. Computer Networks and ISDN Systems, 25(4-
5), pp.454–459. Available at:
http://www.sciencedirect.com/science/article/pii/016975529290039S.
Bertot, J.C., Jaeger, P.T. & Hansen, D., 2012. The impact of polices on government social
media usage: Issues, challenges, and recommendations. Government Information
Quarterly, 29(1), pp.30–40. Available at:
http://www.sciencedirect.com/science/article/pii/S0740624X11000992.
Bikel, D.M. et al., 1997. Nymble. In Proceedings of the fifth conference on Applied natural
language processing -. Morristown, NJ, USA: Association for Computational
Linguistics, pp. 194–201. Available at:
http://dl.acm.org/citation.cfm?id=974557.974586.
Bizer, C. et al., 2009. DBpedia - A crystallization point for the Web of Data. Web Semantics:
Science, Services and Agents on the World Wide Web, 7(3), pp.154–165. Available at:
http://www.sciencedirect.com/science/article/pii/S1570826809000225.
Borthwick, A. & Sterling, J., 1998. NYU: Description of the MENE named entity system as
used in MUC-7. … Conference (MUC-7. Available at:
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.108.6430.
Buitelaar, P. & Ramaka, S., 2005. Unsupervised ontology-based semantic tagging for
knowledge markup. Workshop on learning in web search at 22nd …. Available at:
http://cosco.hiit.fi/search/learninginsearch05/ICMLW4-LWS.pdf#page=34.
Ciravegna, F. et al., 2002. User-system cooperation in document annotation based on
information extraction. Knowledge Engineering and …. Available at:
http://link.springer.com/chapter/10.1007/3-540-45810-7_15.
Cunningham, H. et al., 2001. GATE: an architecture for development of robust HLT
applications. In Proceedings of the 40th Annual Meeting on Association for
Computational Linguistics - ACL ’02. Morristown, NJ, USA: Association for
Computational Linguistics, p. 168. Available at:
http://dl.acm.org/citation.cfm?id=1073083.1073112.
Dalvi, B., Cohen, W. & Callan, J., 2012. Websets: Extracting sets of entities from the web
using unsupervised information extraction. … ACM international conference on Web ….
Available at: http://www.cs.cmu.edu/~bbd/wsdm2012.pdf.
Domingue, J. & Dzbor, M., 2004. Magpie. In Proceedings of the 9th international conference
on Intelligent user interface - IUI ’04. New York, New York, USA: ACM Press, p. 191.
Available at: http://dl.acm.org/citation.cfm?id=964442.964479.
Embley, D.W. et al., 1998. Ontology-based extraction and structuring of information from
data-rich unstructured documents. In Proceedings of the seventh international conference
on Information and knowledge management - CIKM ’98. New York, New York, USA:
ACM Press, pp. 52–59. Available at: http://dl.acm.org/citation.cfm?id=288627.288641.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 78
Title of the Paper
Etzioni, O. et al., 2008. Open information extraction from the web. Communications of the
ACM, 51(12), p.68. Available at:
http://dl.acm.org/ft_gateway.cfm?id=1409378&type=html.
Etzioni, O. et al., 2005. Unsupervised named-entity extraction from the Web: An experimental
study. Artificial Intelligence, 165(1), pp.91–134. Available at:
http://www.sciencedirect.com/science/article/pii/S0004370205000366.
Fader, A., Soderland, S. & Etzioni, O., 2011. Identifying relations for open information
extraction. Proceedings of the Conference on …, pp.1535–1545. Available at:
http://dl.acm.org/citation.cfm?id=2145432.2145596.
Freitag, D. & McCallum, A., 1999. Information extraction with HMMs and shrinkage. … on
machine learning for information extraction. Available at:
http://www.aaai.org/Papers/Workshops/1999/WS-99-11/WS99-11-006.pdf.
Gheorghiu, C. & Nicolescu, R., 2011. SIGMA-SemantIc Government Mash-Up Application:
Using Semantic Web Technologies to Provide Access to Governmental Data. …
(ISPDC), 2011 10th …. Available at:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6108280.
Grishman, R., 1996. Message Understanding Conference-6: A Brief History. Proceedings of
COLING, 96.
Gruber, T.R., 1995. Toward principles for the design of ontologies used for knowledge
sharing? International Journal of Human-Computer Studies, 43(5-6), pp.907–928.
Available at: http://www.sciencedirect.com/science/article/pii/S1071581985710816.
Hayes, P.F.P.-S.P. & Ian, H., OWL Web Ontology Language Semantics and Abstract Syntax.
Available at: http://www.w3.org/TR/owl-semantics/.
Hoffart, J. et al., 2011. Robust disambiguation of named entities in text. , pp.782–792.
Available at: http://dl.acm.org/citation.cfm?id=2145432.2145521.
Hoxha, J. & Brahaj, A., 2011. Open Government Data on the Web: A Semantic Approach.
Emerging Intelligent Data and Web …. Available at:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6076428.
Ji, H. & Grishman, R., 2006. Data selection in semi-supervised learning for name tagging. ,
pp.48–55. Available at: http://dl.acm.org/citation.cfm?id=1641408.1641414.
Johnston, E. & Hansen, D., 2011. Design lessons for smart governance infrastructures.
American Governance. Available at:
http://www.academia.edu/download/30941690/Johnston_Hansen_Gov_3_0_chapter_fina
l.pdf.
Kamegai, S., 2002. Toward ontology-based knowledge extraction from biomedical literature.
GENOME INFORMATICS …, 577(2002), pp.576–577. Available at:
http://jsbi2013.sakura.ne.jp/pdfs/journal1/GIW02/GIW02P078.pdf.
Klischewski, R., 2003. Semantic web for e-government. Electronic Government. Available at:
http://link.springer.com/chapter/10.1007/10929179_52.
Kulkarni, S. et al., 2009. Collective annotation of Wikipedia entities in web text. In
Proceedings of the 15th ACM SIGKDD international conference on Knowledge
discovery and data mining - KDD ’09. New York, New York, USA: ACM Press, p. 457.
Available at: http://dl.acm.org/citation.cfm?id=1557019.1557073.
Lin, T., Etzioni, O. & Fogarty, J., 2009. Identifying interesting assertions from the web. In
Proceeding of the 18th ACM conference on Information and knowledge management -
CIKM ’09. New York, New York, USA: ACM Press, p. 1787. Available at:
http://dl.acm.org/citation.cfm?id=1645953.1646230.
Marshall, C.C. & Shipman, F.M., 2003. Which semantic web? In Proceedings of the
fourteenth ACM conference on Hypertext and hypermedia - HYPERTEXT ’03. New
York, New York, USA: ACM Press, p. 57. Available at:
http://dl.acm.org/citation.cfm?id=900051.900063.
McCallum, A. & Li, W., 2003. Early results for named entity recognition with conditional
random fields, feature induction and web-enhanced lexicons. In Proceedings of the
seventh conference on Natural language learning at HLT-NAACL 2003 -. Morristown,
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 79
Title of the Paper
NJ, USA: Association for Computational Linguistics, pp. 188–191. Available at:
http://dl.acm.org/citation.cfm?id=1119176.1119206.
McGuinness, D.L. & Harmelen, F. van, OWL Web Ontology Language Overview. Available
at: http://www.w3.org/TR/owl-features/.
Medjahed, B., Bouguettaya, A. & Ouzzani, M., 2003. Semantic web enabled e-government
services. Proceedings of the 2003 …. Available at:
http://dl.acm.org/citation.cfm?id=1123287.
Miller, S., Crystal, M. & Fox, H., 1998. Algorithms That Learn To Extract Information BBN:
Description Of The Sift System As Used For MUC-7. … Conference (MUC-7).
Available at: http://aclweb.org/anthology//M/M98/M98-1009.pdf.
Misuraca, G., Broster, D. & Centeno, C., 2012. Digital Europe 2030: Designing scenarios for
ICT in future governance and policy making. Government Information Quarterly, 29,
pp.S121–S131. Available at:
http://www.sciencedirect.com/science/article/pii/S0740624X11000724.
Nadeau, D. & Sekine, S., 2007. A survey of named entity recognition and classification.
Lingvisticae Investigationes, 30(1), pp.3–26. Available at:
http://www.ingentaconnect.com/content/jbp/li/2007/00000030/00000001/art00002.
Navarro-Galindo, J. & Jiménez, J., 2012. Flexible range semantic annotations based on RDFa.
Data Security and Security Data. Available at:
http://link.springer.com/chapter/10.1007/978-3-642-25704-9_14.
Schreiber, R. & Swick, G., 2006. Semantic Web Best Practices and Deployment Working
Group. Available at: http://www.w3.org/2001/sw/BestPractices/.
Sheth, A. et al., 2002. Managing semantic content for the Web. IEEE Internet Computing,
6(4), pp.80–87. Available at:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=1020330.
Skounakis, M., Craven, M. & Ray, S., 2003. Hierarchical hidden markov models for
information extraction. IJCAI. Available at: http://papercut.googlecode.com/hg-
history/98464ac0efb47c55159b313c89b0b305ba1d83f9/PaperCutTesting/targetPDF/succ
ess/hhmm.pdf.
Small, S. & Medsker, L., 2013. Review of information extraction technologies and
applications. Neural Computing and Applications. Available at:
http://link.springer.com/article/10.1007/s00521-013-1516-6.
Soderland, S., 1999. Learning Information Extraction Rules for Semi-Structured and Free
Text. Machine Learning, 34(1-3), pp.233–272. Available at:
http://link.springer.com/article/10.1023/A:1007562322031.
Suchanek, F., Kasneci, G. & Weikum, G., 2007. Yago: a Core of Semantic Knowledge.
Proceedings of the 16th …, p.697. Available at:
http://dl.acm.org/citation.cfm?id=1242572.1242667.
Tim Berners-Lee, 1989. The original proposal of the WWW, HTMLized. Available at:
http://www.w3.org/History/1989/proposal.html.
Tim Berners-Lee, James Hendler & Lassila, O., 2001. The semantic web. Scientific American.
Available at: http://www.scientificamerican.com/article/the-semantic-web/.
Tse-Ming Tsai et al., 2003. Ontology-mediated integration of intranet web services. Computer,
36(10), pp.63–71. Available at:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1236473.
Uschold, M. & Gruninger, M., 2009. Ontologies: principles, methods and applications. The
Knowledge Engineering Review, 11(02), p.93. Available at:
http://journals.cambridge.org/abstract_S0269888900007797.
Vargas-Vera, M. & Motta, E., 2002. MnM: Ontology driven semi-automatic and automatic
support for semantic markup. … : Ontologies and the …. Available at:
http://link.springer.com/chapter/10.1007/3-540-45810-7_34.
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 80
Title of the Paper
Appendices 8
Appendix A, A Public Service RDFa snippet.
<div xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
xmlns="http://www.w3.org/1999/xhtml"
xmlns:foaf="http://xmlns.com/foaf/0.1/"
xmlns:cpsv="http://purl.org/vocab/cpsv#"
xmlns:dct="http://purl.org/dc/terms/"
xmlns:xsd="http://www.w3.org/2001/XMLSchema#"
class="rdf2rdfa">
<div class="description" about="http://www.rsa.ie/org">
<div rel="cpsv:playsRole">
<div class="description"
about="http://cpsv.testproject.eu/id/irl/PublicService/DrivingTest"
typeof="cpsv:PublicService">
<div property="dct:description"
content="Driver testing in Ireland is carried out directly by the
Road Safety Authority to a standard that complies with the EU Directive on Driving
Licences. You can now apply and pay for your driving test online. You will need a
credit card to do this (Visa or Mastercard) or a debit card(Laser) plus a valid e-
mail address. Alternatively, you can download a driving test application form, or
obtain a copy from any motor tax office."
xml:lang="en"/>
<div rel="cpsv:hasChannel"
resource="http://www.rsa.ie/RSA/Learner-Drivers/The-Driving-
Test/Apply-online/"/>
<div rel="cpsv:hasInput"
resource="http://cpsv.testproject.eu/id/irl/Input/MandatoryLessons"/>
<div rel="cpsv:hasInput">
<div class="description"
about="http://cpsv.testproject.eu/id/irl/Input/PPSNumber"
typeof="cpsv:Input">
<div property="dct:title" content="PPS Number" xml:lang="en"/>
</div>
</div>
<div rel="cpsv:hasInput">
<div class="description"
about="http://cpsv.testproject.eu/id/irl/Input/PaymentCard"
typeof="cpsv:Input">
<div property="dct:title" content="Payment Card" xml:lang="en"/>
</div>
</div>
<div rel="cpsv:hasInput">
<div class="description"
t-Gov Workshop ’14 (t-Gov 14)
June 12 – 13 2014, Brunel University, West London, UB8 3PH
Name Surname (First Author) or First Author and Second Authors or First Author et al. 81
Title of the Paper
Appendix B, The RDF/XML representation of the extracted public service and its inputs.
<?xml version="1.0" encoding="utf-8" ?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
xmlns:dct="http://purl.org/dc/terms/"
xmlns:foaf="http://xmlns.com/foaf/0.1/"
xmlns:cpsv="http://purl.org/vocab/cpsv#" >
<rdf:Description
rdf:about="http://cpsv.testproject.eu/id/irl/PublicService/DrivingTest">
<rdf:type rdf:resource="http://purl.org/vocab/cpsv#PublicService" />
<dct:title xml:lang="en">Apply for Driving Test</dct:title>
<dct:description xml:lang="en">Driver testing in Ireland is carried out directly
by the Road Safety Authority to a standard that complies with the EU Directive on
Driving Licences. You can now apply and pay for your driving test online. You will
need a credit card to do this (Visa or Mastercard) or a debit card(Laser) plus a valid
e-mail address. Alternatively, you can download a driving test application form, or
obtain a copy from any motor tax office.</dct:description>
<dct:type rdf:resource="http://id.esd-toolkit.eu/service/1383" />
<foaf:homepage rdf:resource="http://www.rsa.ie/RSA/Learner-Drivers/The-Driving-
Test/Apply-online/" />
<dct:spatial
rdf:resource="http://publications.europa.eu/resource/authority/country/IRL" />
<cpsv:hasInput
rdf:resource="http://cpsv.testproject.eu/id/irl/Input/LearnerPermit" />
<cpsv:hasInput
rdf:resource="http://cpsv.testproject.eu/id/irl/Input/MandatoryLessons" />
<cpsv:hasInput rdf:resource="http://cpsv.testproject.eu/id/irl/Input/DriverNumber"
/>
<cpsv:hasInput rdf:resource="http://cpsv.testproject.eu/id/irl/Input/PPSNumber" />
<cpsv:hasInput rdf:resource="http://cpsv.testproject.eu/id/irl/Input/PaymentCard"
/>
<cpsv:produces
rdf:resource="http://cpsv.testproject.eu/id/irl/Output/DrivingLicenseCard" />
<cpsv:hasChannel rdf:resource="http://www.rsa.ie/RSA/Learner-Drivers/The-Driving-
Test/Apply-online/" />
</rdf:Description>
<rdf:Description rdf:about="http://www.rsa.ie/org">
<cpsv:provides
rdf:resource="http://cpsv.testproject.eu/id/irl/PublicService/DrivingTest" />
<cpsv:playsRole
rdf:resource="http://cpsv.testproject.eu/id/irl/PublicService/DrivingTest" />
</rdf:Description>
<rdf:Description
rdf:about="http://cpsv.testproject.eu/id/irl/PublicService/DrivingTest">
<cpsv:hasInput
rdf:resource="http://cpsv.testproject.eu/id/irl/Input/LearnerPermit" />
</rdf:Description>
<rdf:Description rdf:about="http://cpsv.testproject.eu/id/irl/Input/DriverNumber">
<rdf:type rdf:resource="http://purl.org/vocab/cpsv#Input" />
<dct:title xml:lang="en">Driver Number</dct:title>
</rdf:Description>
<rdf:Description rdf:about="http://cpsv.testproject.eu/id/irl/Input/PPSNumber">
<rdf:type rdf:resource="http://purl.org/vocab/cpsv#Input" />
<dct:title xml:lang="en">PPS Number</dct:title>
</rdf:Description>
<rdf:Description rdf:about="http://cpsv.testproject.eu/id/irl/Input/PaymentCard">
<rdf:type rdf:resource="http://purl.org/vocab/cpsv#Input" />
<dct:title xml:lang="en">Payment Card</dct:title>
</rdf:Description>
</rdf:RDF