In this paper we introduce DiTTO, an online service that allows one to convert an E/R diagram created through the yEd diagram editor into a proper OWL ontology according to three di erent conversion strategies. Ontology design is an activity that involves the use of many di erent languages, resources, and technologies. When dealing with formal or semi-formal languages used by di erent people such as domain experts, knowledge engineers or nal users, a correct transformation strategy can be crucial to guarantee an e ective design. For instance, it seems preferable to adopt intuitive languages when the ontology should be introduced to and/or discussed with an heterogeneous audience, which may not be expert of formal languages and knowledge representation. In these cases, graphical languages seem to support well both ontology understanding and the discussion between knowledge engineers and nal users. The recent project ran by the Italian National Research Council (CNR)4 and SOGEI5, the Information and Communication Technology company owned by the Italian Ministry of Economy and Finance, presents the aforementioned scenario. The aim of the CNR-SOGEI project is twofold. One goal is the reengineering of the E/R (Entity-Relationship) models SOGEI use to describe scal entities (such as taxpayers, laws, deposits, etc.) into standard Semantic Web languages, such as OWL 2. Another goal is to propose tools that facilitate future interactions (changes, re-use in di erent application contexts, etc.) with these semantic models. In this paper we present one of the aforementioned tools called DiTTO, which stands for Diagrams Transformations inTo OWL. As its name suggest, DiTTO is able to translate E/R diagrams expressed in crow's foot notation and created with yEd6, an open source application to quickly and e ectively generate high-quality diagrams, into OWL ontologies.
In Section 2 we present some relevant diagram models originally developed for particular purposes (e.g. software engineering, databases, AI), and progressively adapted to also model OWL ontologies. In Section 3 we introduce DiTTO, describing its implementation and showing how to use it for diagram transformation. We also discuss a set of transformation rules to convert E/R diagrams into OWL. Finally, in Section 4, we present our plans for future developments of the tool, in terms of both new transforming features and diagrams support. 2
One of the most common graphical notations for logical languages is a semantic
network, which can be de ned as a \graphic notation for representing knowledge
in patterns of interconnected nodes and arcs" [
Gasevic et al. [
DiTTO has been implemented as a Web service8, as shown in Fig. 1.
The core of DiTTO is a set of XSLT 2.0 documents included in a Java Web Application Archive (i.e. a WAR le) served as a Tomcat application. These XSLT documents apply several rewriting templates to the source le of the E/R diagram created by means of yEd, which is stored in GraphML format9. 7 http://www.topbraidcomposer.com 8 http://www.essepuntato.it/ditto 9 GraphML speci cation: http://graphml.graphdrawing.org/speci cation.html.
Using the service is quite simple. One needs to specify a URL referring to a yEd diagram, or, alternatively, to choose such a diagram from the le system. Then, after choosing among the available options and after specifying the pre x for naming ontological entities and the full URI of the ontology itself, the \Generate OWL" button can be used to produce an RDF/XML le containing the OWL ontology result of the conversion.
The transformation rules DiTTO implements are as follows. E/R entities, attributes and relations are converted into OWL classes, data properties and object properties respectively. All the subtype relations between E/R entities are converted into rdfs:subClassOf axioms. In addition, DiTTO adds appropriate restrictions to classes according to the edges that link E/R entities. In particular: { a link between an E/R entity and an attribute results in restricting the related OWL class with a quali ed cardinality; { a symbol (i.e. zero-to-many) of a relation results in restricting the domain class with a universal quanti er. This restriction is also added if any of the symbols in the following point is speci ed; { the symbols (i.e. one-to-many), (i.e. one-to-one) and (i.e. zero-toone) result in restricting the domain class with an existential quanti er, a quali ed cardinality and a quali ed maximum cardinality respectively.
In addition to these rules, DiTTO allows one to chochooose what E/R semantics to apply for the transformation. We have identi ed three alternative conversion strategies, which depends on the application of two assumptions: { global semantics (GS) is a characteristic of OWL ontologies (but not typically of E/R), and has the e ect of unifying the formal interpretation of domain and range axioms, property characteristics, and all the restrictions that act at a global level. When GS does not hold, one is not allowed to assume such uni cation, even when the axioms regard two constants with the same name; { unique name assumption (UNA), which is a characteristic in E/R semantics (but not of OWL), and whose consequence is that two objects named di erently always refer to di erent entities in the world.
In particular, the minimal strategy interprets the semantics of E/R in its
purest form (cf. [
Di erent strategies proved useful in the aforementioned project because they address di erent requirements. The minimal strategy is of course conservative with the possible conceptualisations admitted by the original speci cation, while the maximal one allows us to simulate the consequences of E/R into OWL semantics, with its pros and cons, e.g. suggesting domains and ranges, or uni cation of properties, as well as spotting potential issues when applying the simulation.
In Fig. 2 we illustrate a small E/R diagram based on SIOC10, and an excerpt of the OWL ontology returned by DiTTO by using the maximal strategy for the conversion. 10 SIOC Ontology: http://sioc-project.org/ontology.
In this paper we presented DiTTO, an online service that converts E/R diagrams created with the yEd editor into proper OWL ontologies according to three di erent conversion strategies: minimal, intermediate and maximal. Future extensions of the tool will address the management of additional E/R features (such as primary and foreign keys, and multiple and optional attributes), as well as di erent kinds of diagrams (e.g. UML and Gra oo11) as input.