Development of a Controlled Natural Language
Interface for Semantic MediaWiki
Paul R. Smart1 , Jie Bao2, Dave Braines3 and Nigel R. Shadbolt1

1 School of Electronics and Computer Science, University of Southampton, Southampton,
SO17 1BJ, United Kingdom {ps02v, nrs}@ecs.soton.ac.uk
2 Department of Computer Science, Rensselaer Polytechnic Institute, Troy,
NY 12180, USA. baojie@cs.rpi.edu
3 Emerging Technology Services, IBM United Kingdom Ltd, Hursley Park, Winchester,
Hampshire, SO21 2JN, United Kingdom. dave_braines@uk.ibm.com
Abstract. Semantic wikis support the collaborative creation, editing and
utilization of semantically-enriched content, and they may therefore be well-
suited to addressing problems associated with the limited availability of high-
quality online semantic content. Unfortunately, however, many popular
semantic wikis, such as Semantic MediaWiki (SMW), are not sufficiently
expressive to support full-scale ontology authoring. Furthermore, the grounding
of the Semantic Web in formal logic makes both the comprehension and
production of ontological content difficult for many end-users. In order to
address these issues, the expressivity of SMW was extended using a
combination of semantic templates and a Web Ontology Language (OWL)
meta-model. Semantic templates were also used to provide an ontology
verbalization capability for SMW using the Rabbit Controlled Natural
Language (CNL). The resulting system demonstrates how CNL interfaces can
be implemented on top of SMW. The proposed solution introduces no changes
to the underlying functionality of the SMW system, and the use of semantic
templates as an ontology verbalization solution means that end-users can
exploit all the usual features of conventional wiki systems to collaboratively
create new CNL verbalization capabilities.
Keywords: Controlled Natural Languages, Semantic Wikis, Rabbit, Semantic
Web, Ontology Development, OWL
1 Introduction
Ever since the early days of the Semantic Web (SW), there have been concerns about
the usability of semantic technologies for human end-users [1]. The grounding of the
SW in formal logic limits the ability of end-users to participate in the creation,
modification and exploitation of semantic content, and unless such usability concerns
can be addressed, it is possible that the SW will fail to realize its full potential.
Recently, there have been a number of efforts to improve end-users‘ ability to
create and comprehend SW resources by using natural language (NL) interfaces. As
part of their analysis into the kinds of problems encountered by users when working
with the Web Ontology Language (OWL), Rector et al [2] identified the need for a

―pedantic but explicit paraphrase language‖ – a language that would support user
comprehension by substituting (or at least supplementing) formal logic expressions
with NL glosses. Subsequent research has yielded a number of candidate solutions
designed to bridge the SW usability gap via NL interfaces. Aside from the
development of Manchester OWL Syntax [1], and the attempts of some ontology
editing tools to provide NL verbalization solutions [3], members of the Controlled
Natural Language (CNL) community have proposed a number of languages to assist
users with the creation, modification and exploitation of SW resources (see [4] for a
recent review). CNLs are a subset of NL that impose restrictions on both the
generation and interpretation of NL expressions [4]. As a subset of NL, CNLs are
ideally poised to address the SW usability gap: they capitalize on all the
comprehension and productivity benefits of NL, without necessarily undermining the
potential for machine-based processing of SW content.
Usability concerns are, however, not the only problem for the SW community
when it comes to ensuring the future uptake of SW technologies. Further worries
relate to the visibility of practical benefits, the dynamics of conceptual change in
specific communities of interest (see [5]), and the mismatch between end-user
representational requirements and the nature of available ontological content.
Semantic wikis present a potential solution to (at least some of) these problems (see
Section 2). They capitalize on the availability of Web 2.0 technologies (which have
proved very popular in terms of promoting the large-scale participation of user
communities in the generation of online content), and they also avail themselves of
opportunities to collaboratively create, edit and exploit semantically-enriched content.
Given the ability of wikis to support the collaborative creation of online content, and
given the apparent suitability of CNLs as an interface language for the SW, it is
possible that a combination of semantic wikis and CNLs could be used to good effect
in terms of promoting the greater availability of high-quality online semantic content.
In this paper, we present a semantic wiki system that avails itself of CNL-based
ontology verbalization capabilities. The system is called the WikiOnt-CNL system1,
and it is implemented on top of Semantic MediaWiki (SMW) [6], which is one of the
most popular and mature semantic wikis currently available. The CNL we focus on,
for the purposes of the current paper, is the Rabbit language [7, 8], developed by
research staff at the Ordnance Survey2 of Great Britain.
The structure of the paper is as follows: Section 2 describes background research
and ideas relating to the use of both CNLs and semantic wikis; Section 3 provides an
overview of closely related work in this area, specifically the work of Tobias Kuhn to
develop a CNL-enabled wiki system using the Attempto Controlled English (ACE)
CNL [9, 10]; Section 4 describes the technological infrastructure and capabilities of
the WikiOnt-CNL system; Section 5 highlights a number of shortcomings of the
WikiOnt-CNL system; Section 6 provides an overview of future work; and, finally,
Section 7 summarizes the key achievements to date and comments on the potential
implications of this work for future forms of Web-enabled intelligence.

1 See http://tw.rpi.edu/proj/cnl/Main_Page
2 http://www.ordnancesurvey.co.uk/oswebsite/

2 Background: The Semantic Web, Wikis and Controlled Natural
Languages
The SW provides a vision of advanced information search, retrieval and processing,
made possible by the availability of large bodies of distributed, but heavily
interlinked, data [11]. Although the realization of this vision is now possible, thanks
to the efforts of those working in the computer science disciplines, many would argue
that we are still waiting for it to become actual. The problem is that the full benefits of
the SW are at least partially dependent on the availability of high-quality semantic
content, and such content still seems to be in somewhat short supply. In accounting
for this dearth of online semantic content, we encounter a number of potential worries
and concerns. These include:

 Poor usability. SW technologies are often seen as being difficult to learn and use.
This imposes a high entrance barrier on those who might otherwise participate in
the creation of semantic content.
 Conceptual mismatches. There often seems to be a mismatch between the
representational requirements of a specific user community and the kind of
ontologies that are available for use on the SW. The source of this mismatch may
stem from a failure of published ontologies to keep pace with the rate of conceptual
change in a specific domain (see [5]), or it may simply stem from the fact that
ontology engineers are sometimes far removed from the kind of application
domains in which an ontology is ultimately likely to be used.
 Visibility of practical benefits. While the practical benefits of the SW are often
apparent to those who work in the area of Web technologies, such benefits are
often not immediately visible to the wider user community. In particular, it is often
difficult for individuals to see the benefits of semantic technologies, either for
themselves or for the social collectives in which they live and work.
 Rewards and incentives. Finally, there is notion of what we refer to (following
psychological research into instrumental and associative conditioning) as the
reinforcement schedule associated with the use of semantic technologies. The
reinforcement schedule in the case of the SW is, in fact, a complex one; it features
elements of both probabilistic and delayed reward and thus might perhaps be best
characterized as ‗probabilistic delayed reinforcement‘. An alternative (and perhaps
more apt) name might be ‗socially-contingent delayed reinforcement‘, since the
size of the eventual rewards for participating in the SW ultimately depends on the
actions of other social agents. Essentially, the idea is that the rewards for semantic
technology adoption depend, to a large extent, on what everyone else decides to
do; if only a minority of individuals and organizations decide to embrace semantic
technologies, then the full benefits of the SW are unlikely to be realized. The
reinforcement schedule for adopting semantic technologies is thus somewhat
problematic: not only are the rewards temporally displaced, but the actualization of
those rewards depends, to a large extent, on the decisions and actions countenanced
by other agents.

In thinking about how to approach these issues, it is worth considering that Web
users will contribute large quantities of online content under at least some
circumstances. The first wave of the Web (now commonly referred to as Web 1.0)
required users to learn new tools and languages, but many users (even those not well-
versed in information technology or computing) still took up the challenge of creating
and maintaining online content. With the advent of Web 2.0, the process of adding
content became even easier. Web 2.0 websites such as Wikipedia, Flickr, Facebook
and Twitter, are all associated with large quantities of user-generated content3, and,
what is more, in some cases at least, the user-generated content seems to be of
reasonable quality. A recent study of Wikipedia, for example, found that a sample of
Wikipedia pages contained content that was approximately equivalent (in terms of its
accuracy) to that of corresponding articles in the highly respected Encyclopedia
Britannica [12]. What this shows is that some applications can encourage the large-
scale participation of users in the creation and editing of online content, and, for the
most part, much of this content seems to be of reasonable quality.
One question we can now ask is whether the advent of Web 2.0 software (social
software, as it is commonly called) provides us with an effective means of promoting
the emergence of the SW (sometimes referred to as Web 3.0). The general idea is that
by capitalizing on the availability of popular collaborative content-editing systems
(e.g. wikis), we may be able to promote the creation and exploitation of high-quality
semantic content (perhaps bypassing or at least attenuating the impact of some of the
aforementioned barriers to the adoption of SW technology). The following are some
potential benefits of using collaborative content-editing systems, such as wikis:

 Exploitation of specialist expertise, skills and interests. Collaborative
knowledge editing environments make best use of available end-user abilities.
Domain experts can focus on the entry of domain-specific knowledge (often using
unconstrained NL), while knowledge engineers can use the expert-contributed
knowledge to create more formal knowledge structures.
 Collective responsibility. With collaborative editing comes collective
responsibility. By enabling all users to contribute information, the boundary
between content creator and content consumer is effectively blurred. Users are
enabled to became personally involved in the creation and maintenance of
information, information that will ultimately be consumed by both themselves and
other agents. This radically alters the psychological and sociological significance
of online information content.
 Conceptual change and community consensus. As has been pointed out by a
number of commentators (e.g. [13]), ontologies are not just formal representations
of domain knowledge, they are also community contracts about the kind of
knowledge that is deemed to be important in a domain. This emphasis on
community consensus suggests that the current approach to ontology development,
one in which a large number of ontologies is developed by a rather select group of
ontology engineers, is unlikely to be a viable long-term strategy. Instead, we need
to seek a more decentralized and community-based approach to ontology

3 For example, Wikipedia has 3,014,758 English articles as of 27th August 2009. It also has
10,378,900 named user accounts.

development, one in which knowledge structures emerge (deliberately or
otherwise) via the collective actions of (perhaps) large numbers of individuals.
Another benefit of decentralization is that it enables ontologies to evolve in accord
with the conceptual dynamics of a particular user community or epistemic domain.
 Socially-mediated instant gratification. Whereas the conventional approach to
ontology development encounters problems of socially-contingent delayed
reinforcement (see above), the wiki-based approach to creating semantic content
introduces the idea of what we might refer to as ‗socially-mediated instant
gratification‘. This is the idea that the benefits of semantic-enrichment are readily
apparent in a semantic wiki environment. We can use semantic annotations to
(e.g.) perform sophisticated forms of information retrieval or to automatically
assert novel categories of information that are of interest to multiple users (e.g. ―all
articles about movies directed by Spanish film directors who have directed more
than 10 films and who are now deceased‖). What is more, these benefits result
from the collective actions of large numbers of individuals – in other words, they
are socially-mediated. As a result, we encounter a situation in which greater
collective effort yields greater individual and collective rewards.

In spite of all these potential benefits, however, the problem concerning the
usability of SW technologies remains an important concern. The response of the CNL
community to this problem has been to develop a number of CNLs to support
ontology development. Thus, we encounter languages like Rabbit [8], Sydney OWL
Syntax (SOS) [14] and an OWL-compliant subset of ACE, called ACE-OWL [15]),
all of which have been used to develop domain ontologies for the SW. In most cases,
the user evaluation studies that have been undertaken with these languages suggests
that they support the kind of comprehension and production benefits one would
expect to see with NL renditions of formal logic models [8, 16, 17].
All of this leads to a number of specific proposals as to how high-quality semantic
content can be made more available on the SW. One proposal concerns the integration
of semantic technologies with existing Web 2.0 applications. The aim here is to kick-
start the delivery of semantically-enabled (Web 3.0) capabilities by building on the
existing platform provided by Web 2.0 technologies4. A second proposal concerns the
use of CNLs to improve the user experience of creating, consuming, checking and
correcting semantic content. Thus, we end up with a seemingly sensible technology
development strategy: semantically-enable existing Web 2.0 applications and then
extend their user interaction capabilities with one or more CNL interfaces. This is the
aim of the research reported here. Our research aims to show how SMW can be
adapted to support full scale ontology authoring using an OWL meta-model. It also
shows how both ontologies and semantically-annotated wiki content can be serialized
to a specific CNL, namely Rabbit, within the context of the SMW environment.

4 It should probably be borne in mind, of course, that a similar dependency holds between Web
2.0 and Web 1.0: Web 2.0 capabilities are only possible because they build on Web 1.0
technologies (browsers, protocols, markup languages and all the rest).

3 Related Work
There have been a number of attempts to develop tools for creating and editing
ontologies using CNLs (e.g. Ace View [18] and ROO [19]). In the specific context of
applying CNL technology to collaborative Web-based knowledge editing
environments (e.g. wikis), the work that is most closely related to the work reported
here is that associated with the development of the AceWiki system [9, 10]. There are
a number of points of comparison between AceWiki and the current system. The
following list provides an overview of some of the key similarities and differences:

 Wiki system. AceWiki uses a proprietary wiki system that is built specifically to
support content authoring using ACE. WikiOnt-CNL is implemented on top of
SMW, which itself is implemented as an extension to the popular MediaWiki
engine5. WikiOnt-CNL has all the functionality of SMW, and because SMW has
an active developer community there are lots of additional extensions that can be
used to extend its functionality even further (e.g. semantic map extensions enable
users to create, edit and view semantic geospatial data using an inline map view6).
 CNL interface. AceWiki provides an advanced CNL editor for the ACE language
which can be used to create knowledge statements, semantic rules and inline
queries. WikiOnt-CNL does not (at present) provide an editing interface for CNLs;
its functionality is primarily geared towards the verbalization of existing
ontological content using the Rabbit CNL.
 Semantic expressivity. AceWiki can support the representation of models whose
expressivity is greater than OWL, but it does not (at the present time) support the
representation of OWL datatype properties. WikiOnt-CNL provides full support
for OWL via the OWL meta-model extensions referred to in Section 4.1. However,
WikiOnt-CNL does not provide support for models whose expressivity is greater
than OWL (version 1.1)7.
 Rules and reasoning. AceWiki supports the representation of rules and can
perform reasoning on ACE sentences. WikiOnt-CNL can support the
representation of certain types of rules (see [20]), but its reasoning capabilities are
currently limited.
 Support for multiple CNLs. AceWiki, as its name suggests, has been developed
specifically to support the collaborative creation of ontologies using the ACE CNL.
WikiOnt-CNL currently supports the Rabbit CNL, but its functionality can be
extended to accommodate multiple CNLs. Users can create custom CNL
verbalization capabilities by (collaboratively) editing the relevant CNL
verbalization templates (see Section 4.4).
 Export capabilities. AceWiki content can be exported as both ACE texts and
OWL ontologies. WikiOnt-CNL content can be exported in any format providing

5 MediaWiki is the software used by all projects of the Wikimedia Foundation. These projects
include Wikipedia, Wiktionary and Wikispecies.
6 See http://www.mediawiki.org/wiki/Extension:Semantic_Maps.
7 It should also be borne in mind that SMW does not support the full range of simple XML
Schema datatypes. A default installation of SMW includes support for string, Boolean, date,
text, number, URL, and geographic coordinate datatypes.

an appropriate export template exists. Users can use one of the existing CNL
templates, or they can create their own custom export template.
 Import capabilities. AceWiki does not provide an import capability at the present
time. WikiOnt-CNL can import OWL ontologies; however, in some cases, the
CNL texts generated from an imported ontology are of poor quality. This is
because most imported ontologies do not contain the kind of linguistic knowledge
that supports the appropriate morphological realization of ontology elements in
grammatically-diverse sentential contexts (see Section 5).

Of all these features, the one that is perhaps the most notable is the nature of the
editing interface. In the case of AceWiki, a predictive editor interface supports users
in the creation and editing of CNL sentences, so it is well poised to capitalize on the
putative production benefits of CNLs [16]. Our system, in contrast, does not (at the
present time) provide a direct editing interface for CNLs. Rather, it assumes that
semantic content will be created by other means, e.g. the addition of typed hyperlinks
or the creation of ontologies using a custom form-based editor (see Section 4.3).
While the lack of a CNL-based editing interface may seem to be a deficiency of our
approach, it is also important to remember that we are attempting to capitalize on the
existing mechanisms that SMW provides for the creation of semantic content.
Specialist CNL editing interfaces are certainly a desirable addition to the ways in
which users can create and edit ontological content; however, we do not want to
necessarily limit user interaction to these interfaces. Rather, we want such interfaces
to work in concert with existing mechanisms of semantic content creation. AceWiki
does not, at the present time, support the kind of source page editing that is common
to many wiki systems, and it does not, therefore, allow users to add HTML content to
a page and then annotate this content with metadata elements. For us, this is an
important point of difference. We see the conventional forms of wiki-based content
editing as an essential element of the success of wikis in promoting large-scale
collaborative knowledge editing. Firstly, we think that an ability to add unconstrained
NL content to a page is an important aspect of the knowledge engineering lifecycle –
it enables some contributors (e.g. domain experts) to contribute unstructured
information that serves as the source material for more formal knowledge modeling
efforts. This approach enables multiple individuals to contribute to the knowledge
infrastructure of an application in ways that best suit their idiosyncratic skills, abilities
and interests. Secondly, when seen as an attempt to improve the comprehensibility of
semantically-enriched resources, it seems important for CNL-enabled semantic wikis
to provide a number of ways of visualizing and interacting with semantic content.
AceWiki does provide a number of useful capabilities here, including the ability to
add comments to specific CNL sentences, an ability to view inferred as well as
asserted statements, and an ability to create and execute inline queries. What it does
not provide, however, at least at the present time, is an ability to add other types of
content that might prove beneficial to a human end-user‘s understanding of semantic
content. By this we mean multimedia content, page history logs, discussion threads,
and extended narratives written in plain (unconstrained) NL. SMW, as an extension of
the MediaWiki engine, provides all the functionality that is commonly seen in
Wikipedia, and it therefore provides access to all of the aforementioned features.

4 The WikiOnt-CNL System
4.1 Overview
In order to combine the benefits of semantic wikis and CNLs, we extended the
functionality of SMW to provide full OWL 1.1 authoring and Rabbit CNL
verbalization capabilities. The system comprises four types of components, all of
which are built on top of SMW and the MediaWiki engine. The components are:

1. Semantic Templates: Semantic templates provide a means by which the text of a
wiki page can be programmatically generated by the wiki parser engine. In
essence, the wiki parser replaces a template with the text given on the template‘s
source page. The template may be parameterized, such that different kinds of text
output are generated by the parser. Furthermore, templates can be replaced with
another template so that complex forms of template embedding can occur. In the
context of the WikiOnt-CNL system, semantic templates are used to represent
ontological content and generate Rabbit CNL sentences.
2. Semantic Forms: Semantic forms are an extension to MediaWiki that work in
conjunction with semantic templates. Semantic forms enable users to create and
edit structured data using conventional web form controls, e.g. text boxes, list
boxes, checkboxes and so on. In the context of the WikiOnt-CNL system, semantic
forms are used to support the indirect creation and editing of ontology elements
(i.e. classes, properties, individuals) via semantic templates (see Section 4.3 for
more details).
3. Special Pages: Special pages are pages that provide specific services and
functionalities to end-users. Within the WikiOnt-CNL system, special pages are
used to support ontology import/export capabilities, CNL export capabilities and
semantic query capabilities.
4. Built-In Properties: Built-in properties are wiki pages that are defined in the
―Property:‖ namespace. They correspond to relationships defined as part of the
SMW OWL meta-model (see Section 4.2), and they therefore support the
representation of semantic information that surpasses the rather limited semantic
expressivity of a default SMW installation.

The key components of the WikiOnt-CNL system are illustrated in Fig 1. The
following list summarizes the functionality of the WikiOnt-CNL system with specific
reference to these components.

 WikiOnt-CNL Semantic Templates: Semantic Templates form the core of the
WikiOnt-CNL system. They support the generation of Rabbit CNL Sentences, and
they also provide a mechanism to extend the representational expressivity of SMW
using the SMW-mOWL meta-model. In accord with these two functions, WikiOnt-
CNL Semantic Templates come in two basic flavors: SMW-mOWL Templates and
CNL Verbalization Templates. SMW-mOWL Templates encode meta-model
information about the ontology element (i.e. OWL Class, OWL Property or OWL
Individual) described on a particular Wiki Page (this information is ultimately

translated to a set of RDF triples and stored in the SMW database); CNL
Verbalization Templates generate Rabbit CNL Sentences by retrieving semantic
information from the SMW database and structuring the retrieved information
according to the Rabbit syntax specification. Template inclusion and
parameterization are handled by the WikiOnt-CNL Semantic Form components.


Fig. 1. Main components of the WikiOnt-CNL system. Components not shown in the figure
include Special Pages (SPARQL Query, RDF Export, CNL Parser, CNL Export, and Ontology
Import pages) and additional interface components (e.g. CNL Editor).
 SMW-mOWL: SMW-mOWL is a meta-model for OWL ontologies implemented
within the WikiOnt-CNL system. It enables OWL 1.1. ontologies to be developed
within SMW8. SMW-mOWL is implemented within the WikiOnt-CNL system as a
collection of built-in properties that encode information about the various
relationships and semantic axioms associated with specific ontology elements. For
example, the owl:disjointWith axiom is represented in the WikiOnt-CNL system as
the ―Owl:disjointWith‖ property, which is declared in the ―Property:‖ namespace.
Similarly, a logical conjunction of two classes can be represented using the
―Owl:unionOf‖ property, again declared in the ―Property:‖ namespace. Together,
these properties enable the Wiki Pages corresponding to particular Ontology
Elements to be annotated in a way that is compatible with existing SMW
capabilities. Thus, when asserted on the page ―Category:IranianCity‖, the semantic
annotation ―[[Owl:disjointWith::Category:AfghanCity| ]]‖ asserts that the
―IranianCity‖ class is disjoint with the ―AfghanCity‖ class. When parsed by the
Wiki Parser, the annotation will be translated to the triple ―<wiki:IranianCity
property:disjointWith wiki:AfghanCity>‖ and stored in the SMW database. In this
way, information about the relationships and logical axioms associated with

8 By default, SMW provides only limited support for the representation of OWL constructs.
WikiSource
Page
Form Editing
Interface
Semantic Form
Semantic
Template
SMW-mOWL
Template
SMW-mOWL
Component
SMW-mOWL
Rabbit CNL
Verbalization
Template
Rabbit CNL
Sentence
Wiki Page
Category Page Property Page Individual Page
OWL Class OWL Property OWL Individual
Ontology
Ontology
Element
<< edited by
<< generates
<< represents
is meta-model for >>
corresponds to >> corresponds to >> corresponds to >>
Wiki DB
stored in >>
source page >>
<< links to
displayed in >>

Ontology Elements can be represented within the existing annotation framework of
SMW.
 WikiOnt-CNL Semantic Forms: WikiOnt-CNL Semantic Forms support the
creation and editing of Semantic Templates. Because the text in Semantic
Templates ultimately gets replaced by wiki-text, Semantic Templates provide a
convenient way to simplify the creation and modification of complex page content.
Nevertheless, the direct editing of the Semantic Templates included with the
WikiOnt-CNL system is too complex for most casual users; it requires users to
have a detailed knowledge of the kind of templates that are available and the kind
of parameters such templates should be used with. Semantic Forms provide a
solution here. They enable users to interact with a form-based interface that hides
much of the complexity associated with the entry of structured knowledge content.
Within the WikiOnt-CNL system, Semantic Forms are defined within the ―Form:‖
namespace, and they can be invoked from the main interface of a Wiki Page via a
hyperlink. Different Semantic Forms are provided for the creation and subsequent
editing of classes, properties and individuals.
 Wiki Pages: Wiki Pages within the WikiOnt-CNL system correspond to the
standard wiki pages that one sees in typical installations of MediaWiki (e.g.
Wikipedia) and SMW. Because the WikiOnt-CNL system is built on top of SMW
without replacing or eclipsing any of its functionality, a Wiki Page in WikiOnt-
CNL can be treated in exactly the same manner as a Wiki Page in SMW. WikiOnt-
CNL does not oblige end-users to use either the meta-modeling extensions for
ontology authoring; neither does it require end-users to rely on the template-
mediated mechanism for CNL verbalization. In fact, because SMW already
supports the representation of certain types of semantic information, e.g.
taxonomic hierarchies, class membership and relationships between individuals,
the Rabbit CNL Verbalization templates can be used independently of the OWL
meta-modeling solution described here.
 WikiOnt-CNL Special Pages: Special pages within the WikiOnt-CNL system are
used to provide functionalities that go beyond those provided by SMW. Five
Special Pages are being made available as part of the WikiOnt-CNL system. These
include a CNL Parser Page, an OWL Import Page, a CNL Export Page, an
RDF/XML Export Page, and a WikiOnt-CNL SPARQL Endpoint. Of these, the
OWL Import, RDF/XML Export and SPARQL Endpoint pages provide services
that are similar to those already provided by SMW. The reason we have
implemented WikiOnt-CNL-specific variants of these services is because the
existing services do not take account of the semantic enrichment provided by the
SMW-mOWL meta-model. Rather than modify or replace the existing services, we
suggest it is more appropriate to provide new services that specifically cater for
users interested in exploiting the SMW-mOWL extensions.
 CNL Editing Interface: The CNL Editing Interface component constitutes part of
our proposed future work on the WikiOnt-CNL system. (see Section 6). The aim,
in this case, is to develop an editor that provides end-users with the option of
authoring (and editing) ontologies using one or more CNLs.

4.2 OWL Meta-Model Extensions to Semantic MediaWiki
As discussed by Kuhn [10], one drawback associated with some semantic wiki
systems, concerns their rather limited expressivity. The fact is that many semantic
wikis lack the expressivity required for the development and representation of
semantically-expressive ontologies. SMW is no exception here. Its modeling
capabilities are largely limited to subsumption hierarchies and property specifications.
Thus, in order to support the realistic use of SMW as an ontology development
environment, we need to extend its representational capabilities to support the full
range of OWL language elements (such extensions are, of course, also important in
terms of demonstrating the representational capabilities of OWL-compliant CNLs
within SMW).
Within our system, OWL 1.1 authoring capabilities are enabled via the use of
semantic templates and an OWL meta-model, called SMW-mOWL (where the ‗m‘
stands for meta-model). Individual pages within the wiki correspond to a particular
type of ontology element (i.e. class, property or individual), and, within each page,
information about the target ontology element is encoded using a collection of meta-
model oriented semantic templates. The semantic templates follow the structure of the
SMW-mOWL. So, in the case of OWL classes, there are templates to represent both
named and anonymous classes, as well as templates to represent the relationship
between classes (e.g. rdfs:subClassOf and owl:equivalentClass). In the case of OWL
properties there are two templates: one to represent basic information about a property
and one to represent relationships between properties. Finally, in the case of
individuals, there are templates to represent information about class membership and
the relationships between specific individuals.
Wiki pages corresponding to classes are always created in the ―Category:‖
namespace. This accords with the convention in SMW where ―Category:‖ pages are
deemed to represent categories of pages, and the relationship between a category page
and its (direct) sub-categories is interpreted as a subsumption relationship. A
―Category:‖ page can include either the ―Template:NamedClass‖ template or the
―Template:AnonClass‖ template (these represent named and anonymous9 classes
respectively), both of which encode basic information (e.g. text label) about the class.
Clearly, a specific class (anonymous or otherwise) needs to be associated with more
information than just text labels; therefore a ―Category:‖ page often includes multiple
types of SMW-mOWL template (e.g. the ―Template:NamedClassRelation‖ is used to
represent rdfs:subClassOf relationships).
Properties are created within the ―Property:‖ namespace, again following the
convention in SMW. Basic information about a property is encoded using the
―Template:Property‖ template, and relationships between properties (e.g.
rdfs:subPropertyOf) are encoded using a ―Template:PropertyRelation‖ template.
Instances of classes are represented in pages which are in the main namespace of
the wiki (i.e. the page name is not preceded by a colon terminated label, such as
―Category:‖ or ―Property:‖), or in additional namespaces as specified by the wiki
administrator. Relationships between individuals, for example, ―Hamid_Karzai

9 An anonymous class is typically an instance of the owl:Restriction class. It is automatically
assigned a unique number by the WikiOnt-CNL system (e.g. ―Category:-1‖).

isPresidentOf Afghanistan‖ (where ―isPresidentOf‖ represents an instance of the
―Property:IsPresidentOf‖ property), are always associated with the subject page (i.e.
the page corresponding to ―Hamid_Karzai‖ in the case of the current example).
Currently, the wiki system only supports individuals that are instantiated from named
classes (i.e. ―Category:‖ pages containing the ―Template:NamedClass‖ template). No
direct instances of anonymous classes are permitted.
In order to exemplify the use of templates in encoding information about ontology
elements, consider the following class expression, as represented in OWL Abstract
Syntax:
Class(AfghanTribalChief partial TribalChief
restriction(isLeaderOf someValuesFrom(
Tribe restriction(isLocatedIn Afghanistan))))
This class will be represented in the WikiOnt-CNL system using the following set
of wiki pages (the names of the pages are given in bold font; the statements between
curly braces – {{...}} – correspond to specific SMW-mOWL templates).
Category:AfghanTribalChief
{{NamedClass |label=Afghan Chief |plural=Afghan Chiefs }}
{{NamedClassRelation |type=subClassOf |class=TribalChief }}
{{SomeValuesFrom |on property=isLeaderOf |on class=-1 }}
Category:TribalChief
{{NamedClass |label=Tribal Chief |plural=Tribal Chiefs}}
Category:Tribe
{{NamedClass |label=Tribe |plural=Tribes}}
Category:-1 (automatically numbered)
{{AnonClass}} {{AnonClassRelation |type=subClassOf
|class=Tribe}} {{HasValue |on property=isLocatedIn |has
value=Afghanistan}}
Property:isLeaderOf
{{Property |label=is leader of |type=Object}}
Property:isLocatedIn
{{Property |label=is located in |type=Object}}
Afghanistan
{{Individual}}
The semantic templates associated with each page can be created manually, but we
suspect most users will opt to use the form-based interface described in Section 4.3.
As with any template built within the MediaWiki environment, the wiki parser
replaces the text contained within the curly braces by referencing the syntax of the
appropriate template. Thus, if we examine the source of the
―Template:SomeValuesFrom‖ template, we see the following:

[[owl:someValuesFrom::Property:{{{on
property}}};Category:{{{on class}}}|]]
When the ―SomeValuesFrom‖ template on the ―Category:AfghanTribalChief‖
page is parsed by the wiki, it is replaced with the following text:
[[owl:someValuesFrom::Property:isLeaderOf;Category:-1| ]]
This markup asserts a relationship between the AfghanTribalChief class and a text
string, corresponding to ―Property:isLeaderOf;Category:-1‖, via the property
―Property:Owl:someValuesFrom‖ (which is represented as a built-in property of type
‗String‘ in the WikiOnt-CNL system). Based on this mechanism of template-mediated
text substitution, the logical infrastructure of the ontology is encoded in a form that
can be stored in the SMW knowledge base. Importantly, this approach to knowledge
representation within SMW does not require any changes to the underlying SMW
source code or knowledge base; it simply builds on the existing capabilities of SMW
and the MediaWiki engine. Such an approach makes it easy to implement WikiOnt-
CNL in existing SMW systems (the editing and CNL verbalization capabilities can be
installed simply by creating some special purpose ―Property:‖ pages and copying the
semantic templates from the WikiOnt-CNL system to the existing SMW system).
4.3 Ontology Editing Interface

Most users will, we suspect, not relish the prospect of editing the source code of a
wiki page containing SMW-mOWL templates. For this reason, we have implemented
a number of semantic forms that enable end-users to edit the SMW-mOWL templates
using a form-based interface. Semantic forms are an extension to the MediaWiki
engine, and they enable developers to support users in entering structured data. The
availability of such forms provides us with a means to support the creation and editing
of SMW-mOWL templates.
Fig 2. illustrates the editing interface for a specific (named) class (the
AfghanTribalChief class we encountered earlier). We can see from Fig 2. that the
editing interface for classes is composed of multiple sections, each corresponding to a
specific SMW-mOWL template. Thus, the ‗Basic Information‘ section corresponds to
the ―Template:NamedClass‖ template and permits the editing of basic information
associated with a named class; the ‗Relation to other classes‘ section corresponds to
the ―Template:NamedClassRelation‖ template and permits the assertion of (e.g.)
subclass relationships; and, finally, the ‗The class must have some property values
from‘ section corresponds to the ―Template:SomeValuesFrom‖ template and enables
users to assert information about existential restrictions (in this case, that an
―AfghanTribalChief‖ is the leader of at least one thing that is both a Tribe and is
located in Afghanistan). Other sections provide editing interfaces for the other types
of SMW-mOWL template.

Fig. 2. The form-based editing interface associated with OWL classes. The specific class being
edited in this case is the ―Category:AfghanTribalChief‖ class.
4.4 Rabbit CNL Verbalizers
Once a user has created ontology elements, we need to consider how ontological
information is presented in the context of a specific wiki page. There are clearly a
number of options here, but, in accord with the general aim of improving end-user
comprehension of ontological content, we present a CNL verbalization solution
whereby users can view the logical statements associated with an ontology element
using the Rabbit CNL.
The verbalization solution we present is based on the use of semantic templates,
and all the usual advantages of wiki template solutions apply here. Thus, the CNL
verbalization solution can be enabled in any existing or new SMW installation simply
by copying the appropriate templates. Moreover, the end-user community can edit the
templates in any way they see fit, perhaps to address template errors or to update
specific templates in accord with the grammatical changes associated with particular
CNLs. Finally, it is possible for entirely new templates to be created in order to reflect
the introduction of a new CNL specification. The wiki environment supports the
collaborative editing of CNL verbalization templates in the same way that wikis
typically support content editing (i.e. modification of the source page associated with
the wiki article). Moreover, users can exploit the discussion features of the
MediaWiki system to initiate a discussion thread regarding a specific CNL
verbalization template. This enables the template construction/modification effort to
benefit from the collective knowledge of a potentially diverse user community (for
instance, domain experts, linguists, casual end-users and wiki specialists may all

contribute to a discussion about how a specific CNL verbalization component should
be adapted to meet a particular requirement).
All the Rabbit sentences for a particular ontology are rendered using a top-level
CNL verbalization template, called ―Template:CNL.Rabbit‖. This template calls on
other templates to support the sentential serialization of ontological content to the
Rabbit CNL. All the Rabbit CNL templates are stored in the ―Template:‖ namespace,
and specific templates are responsible for the generation of particular kinds of
sentence. Thus, consider the Rabbit sentence: ―Every Pashtun Tribal Chief is a kind of
Tribal Chief‖. This sentence is generated by a specific template, namely the
―Template:CNL.Rabbit.getConceptRelationAssertions‖ template. Part of the mark-up
associated with this template is listed below:
{{#vardefine:label|{{CNL.getLabel|{{{1}}} }}
}}{{#vardefine:super | {{#ask: [[:{{{1|{{FULLPAGENAME}}}}}]]
|?Category= |mainlabel=-|format=list|link=none}} }} {{#if:
{{#var:super}} |{{#arraymap:{{#var:super}}|,|xxx|<li>Every
[[:{{{1}}}{{!}}{{#var:label}}]] is a kind of
[[:xxx|{{CNL.getLabel|xxx}}]]{{{2}}} }}|}}
This template works by retrieving the label associated with the subject of the
sentence using the Template:CNL.getLabel‖ template. This label (―Pashtun Tribal
Chief‖) is assigned to the variable ‗label‘ and the value of the variable is subsequently
used in the phrase ―Every <label> is a kind of‖ (note that Rabbit function words are
highlighted in bold text in the above mark-up). A similar method is used for retrieving
the label associated with the superclass of the subject class (Category:TribalChief in
the current example) and appending it to the Rabbit sentence.
Most templates need to retrieve information from the SMW knowledge base in
order to substitute values into Rabbit sentences. This is accomplished using the SMW
‗#ask‘ parser function, which executes a query against the SMW knowledge base.
Because the structure of the ontology is stored in the knowledge base, the ‗#ask‘
parser function can retrieve information that was previously created using the form-
based interface described in Section 4.3. In addition, because the Rabbit verbalization
templates are not sensitive to the mechanism used for asserting ontological
information, they can also exploit the semantic annotations that are associated with
content represented elsewhere on the wiki page. If, for example, we had a page about
Hamid Karzai in the wiki (i.e. a page with Hamid Karzai as the subject of the page)
and included the following wiki-text on the wiki page‘s source page, then the Rabbit
CNL verbalization templates would generate the sentence ―Hamid Karzai is president
of Afghanistan‖.
As of July 2009, Hamid Karzai is president of
[[IsPresidentOf::Afghanistan]].
At present, all Rabbit verbalization templates are embedded in the top-level SMW-
mOWL templates for particular ontology elements (for example, all the Rabbit
verbalization templates for named OWL classes are embedded in the SMW-mOWL
―Template:NamedClass‖ template). In future work, we aim to make the use of Rabbit
CNL verbalization capabilities more flexible by creating a separate
―Template:RabbitBox‖ template that will render Rabbit sentences wherever it is
embedded on a wiki page.

5 Problems and Issues
As part of our work on the WikiOnt-CNL system, we have implemented a number of
ontologies to test the ontology authoring and CNL verbalization capabilities described
herein. These efforts have revealed a number of problems and issues that we hope to
address in future work. The current section provides an overview of some of the more
important issues.

 Embedded linguistic knowledge and CNL rendering. In English, as with most
NLs, a word can take on various forms when used in a particular sentence. To
borrow an example from Kaljurand [18], the transitive verb ‗to border‘ can be used
in three forms: ―Every country that borders more than 2 countries that border a
country that is bordered by…‖. The morphological realization of these surface
forms depends on access to linguistic knowledge about the base lexical form of an
ontology element (e.g. ―wiki:isLocatedAt‖ has the lexical form ―is located at‖), as
well as knowledge about how inflected variants of the lexical form should be used
in different grammatical contexts. For example, if a property is represented by a
lexical form that is a transitive verb (e.g. ―Property:Contains‖), then it is important
that the third person singular form of the verb (i.e. ―contains‖) is used in sentences
featuring a singular subject, while the third person plural of the verb (i.e.
―contain‖) is used with plural subjects. This basic form of number agreement is
relatively straightforward to implement in the case of transitive verbs, but not all
properties have lexical counterparts that are transitive verbs. Kuhn [21], for
example, identifies four types of property based on their linguistic characteristics:
transitive verbs (e.g. ―constrains‖), transitive verb/adverb combinations (e.g.
―quickly constrains‖), ‗of‘ constructs (e.g. ―part of‖), and adjectives in the
comparative form (e.g. ―larger than‖). The correct rendering of these various
properties in different CNL sentences is not something that can be accomplished
without a fair amount of background linguistic knowledge, so it seems as though
the ontology authoring capabilities in WikiOnt-CNL need to be extended to
include the representation of linguistic knowledge. CNL-based ontology authoring
tools already include support for the representation of this information. Thus,
AceWiki allows users to create properties based on their linguistic category, e.g.
transitive verb, transitive adjective, and so on. Furthermore, in the case of transitive
verbs, the user can add additional linguistic knowledge to the property
specification (i.e. they can specify the third person singular, plural and past
participle forms of the verb). Of course, CNL-based ontology editors are at an
advantage here because their interface is already set-up to support the entry of
linguistically-oriented input. In AceWiki, for example, users indirectly create
ontology elements by specifying words within given linguistic categories.
 Resolving residual comprehension problems. Despite being presented in NL, the
semantics of some CNL sentences can still be difficult for human end-users to
understand. In Rabbit, for example, we encounter the problem of specific sentences
(e.g. ―Relationship has part is transitive.‖) being hard to understand for non-
ontologists [8], while in ACE we see evidence for the usual confusion between the
precise meanings of ‗and‘ and ‗or‘ expressions [17]. As long as the empirical
evidence indicates the possibility for ambiguity or confusion with CNL sentences,

it is important that we take steps to minimize such confusion. One potential
solution to this problem is demonstrated by AceWiki. AceWiki enables content
editors to add comments to specific ACE sentences using plain (unconstrained) NL
(see [22]). Such comments can clarify the meaning of specific sentences in cases
where comprehension problems might arise, and they therefore serve as an
additional comprehension aid to the human end-user. Unfortunately, the
implementation of such a solution in the case of the WikiOnt-CNL system is not
straightforward. This is because CNL sentences are not explicitly represented in
the wiki database; instead, they are dynamically generated each time a wiki page is
loaded. One potential solution strategy here would be to exploit the research
outcomes of CNL user evaluation studies in order to identify the kinds of sentences
that pose potential comprehension problems for human end-users. With this
understanding in place, it might be possible to automatically add comments to
certain types of sentence as and when they are rendered by the wiki system. For
example, if we know that users have difficulty comprehending the meaning of a
Rabbit sentence such as ―Relationship has part is transitive.‖, then we could
automatically create a comment for this sentence that reads as follows: ―For
example, if objectX has part objectY and objectY has part objectZ then objectX
also has part objectZ‖. The mechanism for including these clarification comments
would rely on the same template-mediated text substitution mechanism that is
applied in the case of plain CNL sentences. The difference, in this case, would be
the programmatic identification of potential problem sentences and the subsequent
generation of comments that have been empirically demonstrated to improve end-
user comprehension. Obviously, the implementation of this capability requires
further research on the both the CNL user evaluation and technology development
fronts10.
 Rules and reasoning. SMW has very limited support for reasoning. This is partly
a result of the fact that SMW does not avail itself of particularly expressive
knowledge representation capabilities (in addition the creators of the SMW system
wanted to keep performance overheads to a minimum (see [6]). With the addition
of more expressive modeling capabilities within WikiOnt-CNL, it becomes
possible to implement more sophisticated forms of reasoning, and, in some cases,
the results of such reasoning could be included in a page and rendered as Rabbit
sentences. An initial attempt at implementing reasoning capabilities within SWM
is proposed by Bao et al [20]. Bao et al [20] suggest that a combination of inline
queries, in conjunction with the template mechanism described here, could be used
to support certain forms of rule-based processing with little or no modification of

10 An alternative, and somewhat more straightforward, solution involves the inclusion of a
hyperlink to a separate wiki page (a ―CNL Sentence Page) for each CNL sentence that is
generated within the wiki. Users could use the hyperlink to access the CNL Sentence Page
and add sentence-specific content to the page using standard wiki editing techniques.
Because the name of the CNL Sentence Page would be the same as the CNL sentence it
represents, the CNL Sentence Page could be used to represent the same sentence irrespective
of where it was rendered in the wiki. Moreover, in the manner typical of wiki systems, the
page would only be generated when the user actually created some content, and a visual
indicator (e.g. font colour change) could be used to indicate whether a specific CNL sentence
was associated with user-generated content.

the underlying SMW software. Given that the results of such reasoning processes
can be represented as RDF triples, it seems plausible that a future extended version
of the WikiOnt-CNL system could serialize the results of reasoning processes as
CNL sentences and embed them in specific wiki pages.
 Ontology authoring capabilities. Finally, despite the existence of a form-based
editing interface, it is still rather awkward to create ontologies in WikiOnt-CNL.
The main problem here relates to the fact that complex class expressions need to be
encoded as relationships between multiple pages (often corresponding to
anonymous classes). In order to address this problem, it will be important to
consider the development of integrated CNL (and perhaps advanced graphical)
editing interfaces. These interfaces could support the visualization and editing of
entire ontologies without requiring the user to interact with multiple wiki pages.
6 Future Work
The following are focus areas for future work on the WikiOnt-CNL system.

 Analysis and incorporation of linguistic knowledge. In the context of future
work related to the WikiOnt-CNL system, we plan to undertake a comprehensive
survey of the different kinds of properties used in SW ontologies. This analysis
will help us to understand the linguistic knowledge required for an appropriate
serialization of ontology properties in grammatically-diverse sentential contexts.
The results of this work will guide the development of extensions to both the
SMW-mOWL and the form-based ontology editing interface.
 Implementation and testing of Special Page capabilities. As discussed in
Section 4.1 some of the functionality of the WikiOnt-CNL is realized by Special
Page components. Some of these components require additional work before they
can be fully exploited by users of the system (e.g. the CNL Parser Special Page is
currently incomplete).
 Development of additional CNL verbalizers. WikiOnt-CNL currently provides
verbalization templates for the Rabbit CNL, but we are also planning to develop
templates for SOS and ACE. While we recognize the efforts currently underway to
develop a standard CNL for OWL ontologies11, we suspect that different user
communities may still want to exploit one of the extant CNLs. It may be the case,
for example, that CNLs are differentially suited to certain editing or knowledge
communication functions. If so, then it makes sense for collaborative knowledge
editing environments to accommodate multiple CNLs and be largely agnostic
about the kind of CNL interfaces they make available to end-users.
 Development of a CNL-editing interface. As discussed in Section 5, it is
important to consider the addition of CNL-editing interfaces to the WikiOnt-CNL
system. Ideally, this editing capability should be available to users in addition to
(and not instead of) the other ways of interacting with semantically-enriched
content (see Section 3).

11 See http://code.google.com/p/owl1-1/wiki/OwlCnl.

7 Conclusion
The SW provides a vision of the future potential of the World Wide Web to support
advanced forms of information processing and Web-enabled intelligence. In this
paper, we have discussed our efforts to contribute to the realization of this vision. We
have suggested that a combination of both semantic wikis and CNLs may provide the
right mix of features necessary for large numbers of users to become more involved in
the creation (and exploitation) of online semantic content. The focus of our efforts to
date has been on the highly popular semantic wiki system, SMW, as well as the
Rabbit CNL. We have shown that by using a template-based mechanism, we can
successfully support the representation of highly expressive ontologies within SMW,
with no changes to the underlying software. Furthermore, we have shown that it is
possible to use semantic templates to support the serialization of semantic content
(including standard SMW page annotations) to Rabbit sentences. The particular
advantages of this solution strategy include the open nature of the template-based
mechanism; i.e. the CNL verbalization templates can be collaboratively edited in the
same kind of way as any other wiki page. Members of the end-user community can
use the collaboration features of the host wiki system to adapt CNL verbalization
capabilities to suit community-specific requirements. They can also extend the CNL
verbalization capabilities so as to support the generation of alternative CNLs.
Our future work on the WikiOnt-CNL system seeks to enhance its current
functionality. In addition to the implementation of an embedded CNL editor, we
propose to extend the verbalization capabilities by using embedded linguistic
knowledge. Such features will, we suggest, provide us with a technological platform
from which to assess the effect of user-friendly collaborative knowledge editing
systems on the general availability of high-quality semantically-enriched content.

Acknowledgments. This research was sponsored by the US Army Research
laboratory and the UK Ministry of Defence and was accomplished under Agreement
Number W911NF-06-3-0001. The views and conclusions contained in this document
are those of the authors and should not be interpreted as representing the official
policies, either expressed or implied, of the US Army Research Laboratory, the U.S.
Government, the UK Ministry of Defence, or the UK Government. The US and UK
Governments are authorized to reproduce and distribute reprints for Government
purposes notwithstanding any copyright notation hereon.
References
1. Horridge, M., Drummond, N., Goodwin, J., Rector, A., Stevens, R., Wang, H.H.: The
Manchester OWL Syntax. OWL Experiences and Directions Workshop (OWLED‘06) at
ISWC2006, Athens, Georgia, USA (2006)
2. Rector, A., Drummond, N., Horridge, M., Rogers, J., Knublauch, H., Stevens, R., Wang, H.,
Wroe, C.: OWL Pizzas: Practical Experience of Teaching OWL-DL: Common Errors &
Common Patterns. European Conference on Knowledge Acquisition (EKAW-2004),
Whittlebury, UK (2004)

3. Hewlett, D., Kalyanpur, A., Kovlovski, V., Halaschek-Wiener, C.: Effective Natural
Language Paraphrasing of Ontologies on the Semantic Web. End User Semantic Web
Interaction Workshop, Galway, Ireland (2005)
4. Smart, P.R.: Controlled Natural Languages and the Semantic Web, Technical Report,
School of Electronics and Computer Science, University of Southampton, Southampton,
England (2008)
5. Smart, P.R., Engelbrecht, P.C.: An Analysis of the Origins of Ontology Mismatches on the
Semantic Web. 16th International Conference on Knowledge Engineering and Knowledge
Management (EKAW'08), Acitrezza, Catania, Italy (2008)
6. Krötzsch, M., Vrandecic, D., Völkel, M., Haller, H., Studer, R.: Semantic Wikipedia.
Journal of Web Semantics 5, 251-261 (2007)
7. Dolbear, C., Hart, G., Kovacs, K., Goodwin, J., Zhou, S.: The Rabbit Language:
Description, Syntax and Conversion to OWL, Technical Report, Ordnance Survey,
Southampton, UK (2007)
8. Hart, G., Johnson, M., Dolbear, C.: Rabbit: Developing a Control Natural Language for
Authoring Ontologies. 5th European Semantic Web Conference (ESWC'08), Tenerife, Spain
(2008)
9. Kuhn, T.: AceWiki: A Natural and Expressive Semantic Wiki. Semantic Web User
Interaction Workshop at CHI 2008, Florence, Italy (2008)
10. Kuhn, T.: AceWiki: Collaborative Ontology Management in Controlled Natural Language.
3rd Semantic Wiki Workshop hosted by the 5th European Semantic Web Conference
(ESWC'08), Tenerife, Spain (2008)
11. Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American 284, 34-
43 (2001)
12. Giles, J.: Internet encyclopaedias go head to head. Nature 438, 900-901 (2005)
13. Hepp, M., Bachlechner, D., Siorpaes, K.: OntoWiki: community-driven ontology
engineering and ontology usage based on Wikis. International Symposium on Wikis,
Odense, Denmark (2006)
14. Cregan, A., Schwitter, R., Meyer, T.: Sydney OWL Syntax—towards a Controlled Natural
Language Syntax for OWL 1.1. OWL Experiences and Directions Workshop, Innsbruck,
Austria (2007)
15. Kaljurand, K., Fuchs, N.E.: Bidirectional mapping between OWL DL and Attempto
Controlled English. 4th Workshop on Principles and Practice of Semantic Web Reasoning,
Budva, Montenegro (2006)
16. Engelbrecht, P.C., Hart, G., Dolbear, C.: Talking Rabbit: a User Evaluation of Sentence
Production. Workshop on Controlled Natural Language (CNL'09), Marettimo Island, Italy
(2009)
17. Kuhn, T.: How to Evaluate Controlled Natural Languages. Workshop on Controlled Natural
Language (CNL'09), Marettimo Island, Italy (2009)
18. Kaljurand, K.: ACE View - An ontology and rule editor based on Attempto Controlled
English. 5th OWL Experiences and Directions Workshop (OWLED'09), Karlsruhe,
Germany (2008)
19. Denaux, R., Dimitrova, V., Cohn, A., Dolbear, C., Hart, G.: Rabbit to OWL: Ontology
Authoring with a CNL-based Tool. Workshop on Controlled Natural Language (CNL'09),
Marettimo Island, Italy (2009)
20. Bao, J., Ding, L., Smart, P.R., Braines, D., Jones, G.: Rule Modeling Using Semantic
MediaWiki. 3rd Annual Conference of the International Technology Alliance (ACITA'09),
Maryland, USA (2009)
21. Kuhn, T.: Attempto Controlled English as an Ontology Language. Reasoning on the Web
with Rules and Semantics, Munich, Germany (2006)
22. Kuhn, T.: How Controlled English can Improve Semantic Wikis. 4th Workshop on
Semantic Wikis, Heraklion, Greece (2009)