Integrated ontology development methodology

A requirements-driven approach for seamless ontology development

Introduction

In this post we will take a look at a methodology for ontology development as an integrated approach. The above graphic provides a high level view of how to conduct ontology development with maximum coverage and places the core idea of requirements management at the heart of it. The methodology draws from several strands of modelling approaches used in enterprise analysis and architecture, as well as observation patterns made over several applied ontology projects. It is to be noted that the methodology does not make any assumptions on the timescales involved for completing each phase and it can be applied in both agile and waterfall settings. Furthermore, it is completely independent of any tools and domain to be modelled and, therefore, has wide-ranging outreach.

Requirements management

Requirements Management provides the ‘glue’ across all the phases of the integrated ontology development methodology. This ensures traceability throughout execution of the methodology so that the inter-dependencies across phases are factored into the ontology development cycle. This way, it becomes possible to avoid an ‘over-the-wall’ approach but instead concentrate on ontology development as a traceable and iterative process where knowledge can be exploited collaboratively.

Goal and scope definition

This phase is regarded as the starting point of any ontology development cycle. It is largely a preparatory phase concerned with the identification of the domain and subject area being studied. It also lists the aims and objectives as well as the high level specifications, which not only guide the subsequent phases but are also later used for evaluating the degree of success of the ontology development activity. The scope of the ontology to be developed is defined in terms of its expected boundaries, coverage and assumptions. The Goal and Scope Definition phase may consider historical information available at hand to accelerate the ontology development process, such as existing ontologies and previous lessons learnt information. Overall, the high level view provided by this phase helps to encompass the requirements for the subsequent phases of the methodology.

Information gathering and elicitation

This phase helps to achieve a deeper understanding of the domain under study. Useful methods to conduct this phase are, for example, through workshops and interviews with subject matter experts, the analysis of existing content and the extraction of information from relevant sources. Furthermore, because this phase helps raise further awareness of the domain under study, it is sometimes necessary to revise and refine the original Goal and Scope Definition phase. In the Information Gathering and Elicitation phase, it may be required to do some very early and rough-cut organisation of the gathered information, where no direct information structuring work is performed. It is also customary to consider emerging patterns in the way that certain concepts and ideas start to unfold.

Initial structuring

This phase is about making sense out of the gathered information and using information organisation methods for finding trends, rationalising and structuring information. Methods used are, for instance, through collation of a term pool that captures concepts, short phrases and what they mean. Refinement of that information leads to finding out what works as potential classes, relations and logic for an ontology. Things like informal definitions of concepts, pseudo-logic and natural language descriptions are written down for subsequent review and use in ontology documentation. Moreover, it may be necessary to share and review the initial structures with both experts and non-experts. For that purpose, diagrammatic representations of the initial ontology structures are extremely handy.

Formalisation

This phase involves the use of a suitable ontology development environment (ontology editor, rule engine, etc.) to encode, refine and test the initial structures as a formal model expressed in a chosen knowledge representation formalism. Classes, relations, logic, etc. are captured. Methods used to ensure that the formalisation is correct are, for example, through partial instantiation of the ontology, classification, consistency and completeness checks, as well as other tests that are relevant to verifying the intended behaviour of the ontology structures.

Deployment

This phase is concerned with ontology publishing and release as well as scaling into an ontology-driven system. In the Deployment phase, knowledge graphs are built and any work necessary for fully instantiating, integrating and interfacing the ontology is performed and tested. In practice, this phase can lead to, for example, proof of concept/value, enterprise ontology roll-out, user acceptance testing of an ontology-driven system, among others.

Evaluation

This phase is where the evaluation of the developed ontology and its deployment is conducted. It requires looking back at the Goal and Scope Definition phase and assessing the extent to which the aims and objectives have been fulfilled and requirements met, in the context of the established scope. Several methods can be used for evaluation, where these fall broadly as qualitative and quantitative analysis methods. In the context of the methodology, ontology development is viewed as a knowledge management activity and for this reason, it is important to conduct, for example, lessons learnt reviews for continuous improvement purposes. This then feeds back into the Goal and Scope Definition phase to meet the needs of future ontology development cycles.