1. Introduction

Despite early uses of the term ontology^[1][2] to designate a “theory of a modeled world”,^[3] the first formulation of a definition for ontology in the context of information systems it only happened in 1991: “the ontology of a system consists of its vocabulary and a set of constraints on the way terms can be combined to model a domain.”^[4]. Some authors, e.g.^[5], consider the study of Neches and others to be a pioneer in the area of ​​Information Science (IS) despite the connection of its authors, particularly Gruber, to the area of Artificial Intelligence (AI). Others authors, e.g.^[6][7], pointing to an assumed author of IS^[8] as the first author of the area to address ontology.

Different views of IS, with a more or less computerized focus^[9], contribute to the lack of terminological clarity given the use of the same term, ontology, to designate related but distinct artifacts as to their objectives. Prima facie, while in IS the emphasis is placed on building structures for document content representation and retrieval in Computer Science (CS) the intention is to create models of the world emphasizing the process of automated inference^[10]. Directly associated with the specificity of the objectives of the two areas, some authors, e.g.^[11], place the fields of study: Knowledge Organization (KO) and Knowledge Representation (KR), respectively, within IS and within CS. Although the distinction of these areas of study may be necessary, in the development of knowledge organization systems (KOS) it does not seem possible to separate the two processes. The organization of knowledge is a condition for its representation, which in turn functions as an instrument for the organization to be effective^[12].

The set of items called KOS is vast and unclear as to the meaning of much of the terminology used for the various types where the term ontology is paradigmatic of this ambiguity^[13][14][15]. The disregard of the multiple dimensions (intrinsic and extrinsic) of the various types of KOS and the undifferentiation between ideal KOS types and particular instances are pointed out by Souza and others^[14] as causes for this lack of clarity. Considering only the modeling process, as the authors point out: “[i]t might be asserted that all KOS are the products of some kind of ontological modeling, but using the term ‘ontologies’ arbitrarily can cause confusion.”^[14] Applying a similar criterion, we could also designate as ‘classifications’ all KOS as they employ in their development the process of classifying, as Soergel^[16] appears to do: “other fields, such as AI, natural language processing, and software engineering, have discovered the need for classification, leading to the rise of what these fields call ontologies.” Disagreeing with Soergel's position, Hjørland^[17] considers ontologies "more general and more abstract forms of KOS" than traditional ones, such as classification systems and thesauri, which, for the author, can be understood as "restricted kinds of ontologies."

The fact that many researchers, particularly in the CS-related community as point out in^[18], adopt Gruber’s definition: “[a]n ontology is an explicit specification of a conceptualization,” proposed in^[19] and reinforced in^[20], favors the indiscriminate use of the term ontology to designate different types of KOS. Criticizing this “mechanical definition” and the uniquely technical guidelines of the W3C RDFS and OWL standards, Kless and others^[21] state: “In fact, it implies that any KOS could simply become an ontology by simply changing its representation format.” The authors distinguish two types of ontologies: “data modeling ontologies,” tied to the RDF-based semantics; and the “reality representation ontologies” associated with the description logic semantics. Adding, in relation to the latter: “[i]t can be questioned, whether they should be called ontologies at all and whether they are truly interoperable in the sense of being combinable and reusable.”^[21]. Since the 1990s, the proliferation of these systems was due to the search to provide the Web with systems capable of automatic inferences in the context of the so-called Semantic Web (SW). The ability to generate inferences is, however, restricted to the semantic expressiveness of the language used which, in the case of RDFS, is quite limited. RDFS is a web-oriented language and not a true SW (or KR) language^[22]. Despite these limitations, it is these systems that underlie what may be called the Web of Linked Data^[23] and it is possible to see in this development of the Web a “democratization” of knowledge representation^[24].

In terms of interoperability and reusability these lightweight ontologies, as they are often called, face also several limitations due to the ad-hoc approach generally employed in their construction^[25]. One consequence of the ad-hoc approach is the continuation of the siloing of data, which ontologies should solve or at least minimize. That's one of the reasons Smith^[26] considers this the wrong way to build ontologies. While condescending that a similar approach may be used in some cases of purpose-built application ontologies, the researcher totally dismisses it for the so-called reference ontologies. These, designed to serve scientific purposes, should be developed according to the “principle of orthogonality”^[26]. This approach would not only address the data silos problem, but would bring additional benefits such as: mutual consistency of ontologies; unnecessary mapping between ontologies; reduced redundancy; facilitated findability of specific ontological resources; and optimized management of the ontological labor division. Essential to Smith's approach is the articulation with ontological study from his disciplinary area of origin - the philosophy: “information-systems ontology is itself an enormous new field of practical application that is crying out to be explored by the methods of rigorous philosophy.”^[27] Articulation seen as necessary for the correct development of ontologies by various researchers, e.g.^[10][28][29].

The rigor sought in the ontological representation of reality, advocated for systems of the type of reference ontologies, is relegated to the background by authors such as Gruber^[30]: “[i]f ontologies are engineered things, then we don’t have to worry so much about whether they are right and get on with the business of building them to do something useful.” It is, like Smith^[26] says, “as if all ontologies, both inside and outside science, are assigned by default the status of application ontologies.” Linked to this approach to ontologies is the conception that ontologies are the result of the common agreement of a community over a portion of the world. This conception is questioned by Poli and Obrst^[28] since this result is usually obtained by the lowest common denominator whose utility will be quite doubtful “because it is inconsistent, has uneven and wrong levels of granularity, and doesn’t capture real semantic variances that are crucial for adoption by members of a community.”^[28]

2. Conclusions

Summarizing, two perspectives on ontologies coexist in KOS ecosystems, on the one hand, we have ontologies viewed as a more complex conceptual system, on the other, a system that organizes ontological rather than epistemological knowledge. The focus of ontological analysis is the item to model and not the intentions that motivate the construction of ontology. This process involves an analytical complexity that makes the development of such systems quite onerous. However, it is the quality of this analysis that determines its true usefulness.

Despite the inherent epistemological interference, i.e., the adjudication of the truth by the modeler, it is the objects, relations, and rules of reality that the ontologist must model. Ultimately "[w]ithout ontology, there is no firm basis for epistemology."^[28]