are grounded. The type of meaning understanding can considerably impact on howKOS are
designed and implemented. However, such theoretical and foundational arguments are
usually outside the field of interest and competence of thesaurus specialists (Hjørland, 2016),
who tend to uncritically endorse the received framework.
This paper surveys the relations of KOS, focusing particularly on the generic relation.
The scrutiny of the latter’s underlying assumptions will offer the opportunity to inspect the
notion of a priori with respect to these relations. It will be examined the tree-like hierarchical
image, which has always constituted the dominant model of classification (at least in the
West) and even today inspires logicist approaches to information organization. These
approaches will be compared to hermeneutically oriented approaches, which recognize the
situated nature of knowledge and understand meaning in its complexity and contextuality.
In its final part, the paper’s aim is to elucidate why different types of relations (for instance,
genus-species relations vs perspective hierarchical relations) can play different roles in
information organization, despite the nature of all relations is the same. It will be shown
what this implies with respect to the interoperability issue.
2. The relational structure of thesauri
KOS such as thesauri are tools designed for improving IR. They are based on a natural
language that is turned into a normalized language, where the terms are basically
monosemic and the relations among them are made explicit.
Two types of semantic devices are employed (Svenonius, 2000): one that concerns
basically the linguistic level, which is based on methods for limiting the meanings or
referents of terms; homonyms and polysemes are disambiguated to improve precision in IR,
and one that concerns basically the conceptual level, which provides a structure capable of
enhancing information recall performance, helps improve precision by suggesting more
specific terms to refine the search and helps eliminate unwanted information. Through this
same structure, a representation of the meaning of each term is also given, along with a
portrayal of how one understands a particular subject field (Soergel, 1995).
Two types of interterm relations are distinguished in thesaurus standards such as ISO 25964-
1-2011 (International Organization for Standardization (ISO), 2011): syntactical or a posteriori
relations, which are regarded as document-dependent –an example is the interrelationship
between “offices,”“printers,”and “London”in a work concerning “printers in offices in
London”–and apriorirelations, which are instead seen as document-independent since “they are
generally recognized and could be established through refer ence to standard works, such as
dictionaries and encyclopaedias”(ISO, 2011, p. 17). A thesaurus is comprised of only the latter
type of relations.
The traditional thesaurus format, which was created to address the information needs of
the library and archival fields, includes three overall types of relations: hierarchical
relations, associative relations, and relations of equivalence. Technological advances,
however, are changing the setting in which KOS work, and this calls for reassessing
whether the traditional format still copes with the current needs of information organization.
Many believe that it lacks a well-defined semantics to address these needs (see e.g. Soergel
et al., 2004; Tudhope et al., 2001). A richer and more refined structure is advocated to
enhance thesaurus suitability for artificial intelligence and Semantic Web applications –
higher expressive capabilities are required to allow inference –and to increase the
possibilities for IR and interoperability among different KOS.
In addition, such a refinement is advised for improving the degree of internal structural
consistency. In fact, basic thesaural relations have not always been established consistently,
with the result of producing unpredictable semantic structures (Dextre Clarke, 2001).
In particular, this concerns the hierarchical relation, above all the generic type, whose sound
implementation plays a key role in ensuring the quality of a structured vocabulary. Many