Conceptual analysis and knowledge representation often require to develop an ontological support. The activity of developing a domain ontology is therefore one of the fundamental steps to be carried out when developing a shared model of the knowledge possessed by an organization, and consequently, one of the pilasters of knowledge management.
It is clear, however, that this activity, as well as all other engineering activities, is constrained to be methodologically coherent in its phases, because the control over correctness of the development method is the only possible way to keep events of such an activity on its own track. Though it would seem attractive to deploy software applications which provide automation to the development of such models, the results of automatic ontology tools in the real practice of knowledge management are definitely disappointing.
We thus would prefer to look at formalized methodologies, and in particular, we are interested in methodologies which work from scratch, namely which aim at providing the ontological structure without any general and supposedly shared reference model in front. That choice is not the ideal one, since in general, scalability is a very desirable property, but since the two problems of developing from scratch and integrating existing parts are deeply different we have to commit to one focus.
After having reviewed the existing methodologies, we understood that these methods, which are depicted at different levels of depth in the current literature, are indeed covering in a reasonable way the several aspects of the development of a domain ontology from scratch, but each has both advantages and drawbacks. We thus carried out an analysis, which is the focus of this paper, about the conditions which make the knowledge engineer commit to one single methodology, defining a set of choosing criteria.
We define a meta-methodology which is based upon the analysis of the choosing conditions in several steps, and model this in the specific case of knowledge management activities.
The paper is organized as follows: Sections 2 and 1 are devoted to describe the framework we refer the ontologies to belong, and Section 3 reviews existing methodologies which have been documented in the recent literature. In Section 4 we provide a description of the meta-methodology we have developed. Section 5 takes some conclusions and sketches further work.