=Paper=
{{Paper
|id=None
|storemode=property
|title=Ranganathans Canons Applied to Ontology Engineering: a Sample Application Scenario in Biomedical ontologies
|pdfUrl=https://ceur-ws.org/Vol-776/ontobras-most2011_paper5.pdf
|volume=Vol-776
|dblpUrl=https://dblp.org/rec/conf/ontobras/CamposCC11
}}
==Ranganathans Canons Applied to Ontology Engineering: a Sample Application Scenario in Biomedical ontologies==
https://ceur-ws.org/Vol-776/ontobras-most2011_paper5.pdf
Ranganathans Canons applied to ontology engineering:
a sample application scenario in biomedical ontologies
Linair Maria Campos1, Maria Luiza de Almeida Campos2,
Maria Luiza Machado Campos3
1
Universidade Federal do Rio de Janeiro – CISI/COPPE
Av. Horacio Macedo S/N - Ilha do Fundão – Bloco H-201- Rio de Janeiro - RJ
2
Universidade Federal Fluminense - PPGCI/UFF
Rua Tiradentes 148 - Ingá - Niterói - Rio de Janeiro - RJ
3
Universidade Federal do Rio de Janeiro - PPGI- IM/NCE
Av. Athos da Silveira Ramos, 274 - Ilha do Fundão - Rio de Janeiro - RJ
linair@cisi.coppe.ufrj.br, maria.almeida@pq.cnpq.br, mluiza@ufrj.br
Abstract. Ranganathan, an Indian mathematician and librarian, has proposed
a set of comprehensible canons to provide guidance to the process of building
concept hierarchies. It is our proposition that Ranganathans canons can
contribute to fulfill the gap between the high-level domain conceptualization
guided by top level ontologies and the classification of such concepts within
facets, needed when building ontologies taxonomical structures. In order to
show the utility of Ranganathans canons applied to ontology structuring, we
have analyzed the structure of a biomedical ontology: Gene Ontology (GO).
As result, we have found that many of the existing inconsistencies on GO
hierarchies could be avoided if Ranganathans canons were adopted.
1. Introduction
Ontologies have been increasingly used since the early 90s, especially in complex
domains such as Biomedicine, where the multitude of concepts and the need to deal
computationally with resources described by them, has urged the adoption of standard
vocabularies. However, the fast growing nature of the body of knowledge being
described and the necessity of fast solution to the issue of concepts standardization has
given rise to vocabularies such as the Gene Ontology [Gene Ontology Consortium,
2001], which has been created without a sound methodology and has been largely
adopted as a de facto standard, despite its many structural problems, as mentioned in
literature [Smith e Kumar, 2004][Smith, Williams, e Schulze-Kremer, 2003], which
affects its efficient utilization.
Notions underlying concepts nature, materialized in classes of top level
ontologies, have given significant contribution to ontology structuring, as it allows
domain concepts to be identified and grouped together in basic categories according to
pre-defined basic features. These top level classes are usually chosen according to
principles discussed in areas such as Philosophy, Cognitive Sciences and Psychology,
providing a sound basis for identifying the nature of concepts in a less ambiguous way.
However, once groups of concepts with the same nature are identified, there is still the
61
need to organize them in arrays (horizontal series of sibling concepts) and chains
(vertical series of concepts, organized hierarchically): this is one of the challenges of
ontology structuring.
Underpinned by more than fifty years of hands on experience in information
classification and structuring of big vocabularies, Shialy Rammarita Ranganathan, an
Indian mathematician and librarian, has proposed a set of principles, or canons,
[Ranganathan,1951;1963;1967a, 1967b], tailored to provide guidance to the process of
working on the level of ideas. This level is the space where the concepts of a given
domain are organized, building a system of concepts [Campos e Gomes, 2008]. These
canons were meant to be used for bibliographic classification, in the context of the
development of documentary languages with taxonomic structures such as thesauri and
controlled vocabularies.
It is our proposition that Ranganathans’ canons can contribute to fulfill the gap
between the high-level domain conceptualization guided by top level ontologies and the
systematic classification of such concepts within facets, needed when building
ontologies taxonomical structures. In order to show the utility of these canons applied to
ontology engineering, we have applied a set of those canons to the widely adopted, de
facto standard, Gene Ontology (GO) [Gene Ontology Consortium, 2001].
As a result, we have observed the timeliness and relevance of his work, as a
series of existing inconsistencies on GO hierarchies, could be avoided if Ranganathans
canons were adopted.
The remainder of this article is structured as follows: in section 2 we present
related work. In section 3 we discuss Ranganathans’ canons. In section 4 we analyze the
Molecular Function branch of GO in accordance with Ranaganathans canons. Finally, in
section 5 we present our conclusions.
2. Related work
Ontology structuring has impact on knowledge reasoning, especially when based on
subsumption relations. Reasoners expect ontologies to comply with certain rules of
classification, and ill-formed hierarchies can lead to false results. On the other hand,
implicit structuring strategies used to form subhierachies can hinder human
comprehension of ontology classification rationale leading to ambiguity when using and
extending ontologies.
In order to tackle those issues, researchers such as Guarino and Welty [2004]
have proposed the use of philosophical notions, such as rigidity, identity, and unity,
materialized on top ontologies, to guide the identification of concepts nature and, thus,
to provide foundational principles to evaluate the conceptual correctness of
specialization relationships [Guizzardi, 2005]. In this sense, Guarino and Welty have
used those notions to underpin the OntoClean methodology [Guarino & Welty, 2004,
2002a, 2002b] aiming at building “clean” taxonomical structures on ontologies. Inspired
by the work of Guarino and Welty, Guizzardi also has proposed a theory presenting a set
of postulates aimed to aid the construction of well grounded conceptual models
[Guizzardi, 2005; Guizzardi, Wagener and Sinderen, 2004]. The idea behind these
approaches is to axiomatize a set of rules that can be applied systematically to
taxonomies and on doing so, prevent structural errors. For example, based on the axiom
62
that rigid concepts cannot be subsumed by anti-rigid concepts, the concept human
(rigid) cannot be subsumed by the concept student (anti-rigid).
Smith (2005) observes, complementary, the role of definitions to identify
attributes “in a consistent manner, thus assuring their transitive inheritance through a
type hierarchy”, and points that the definition of a concept within an ontology should
encompass the definition of all its parents. Besides, all intermediate classes in the
hierarchy where the concept is situated should also be defined, in order to ensure
transitive inheritance of essential characteristics. With the intention of establishing
guidance for building well formed hierarchies, built according to a sound classification
systematic, Smith proposes a set of axioms for a “Theory of Biological Classification”.
According to Smith (2005), those axioms were motivated by the theory of classes found
in Aristotle’s writings. As an example, we can mention the axiom that addresses the
issue of polihierarchy and states that a species should never have two parents:
In Information Science, Dahlberg (1978a, 1978b) also stresses the importance of
definitions as they make explicit the contents of concepts and provide the elements that
forge the relationships between them. Dahlberg, through her Concepts Theory, proposes
also that definitions reveal a set of common characteristics which are useful to build any
system of classification or thesaurus [Dahlberg, 1983]. Dahlberg, however, focuses on
proposing principles to organize concepts in broad categories, and, although
highlighting with examples the importance of definitions when structuring hierarchies,
she does not provide detailed guidance on how to organize them systematically in
subclasses.
To exemplify the problems caused by the lack of a systematic approach on
structuring an ontology taxonomy, Smith points out several issues in the Gene Ontology
(GO) hierarchies. Although his axioms can help to identify solutions to those problems,
members of the GO community were not very receptive to the proposal, perhaps due to
the complexity that comes with it: “(…) When challenged with such problems, the
members of the GO and associated communities standardly insist that their concerns are
those of practicing biologists, and that they are thus not concerned with the sorts of
scrupulousness that are important in logic” [Smith, 2005].
Guarino and Weltys’ (2004) as well as Smiths’ (2005) proposals have the focus
on identifying concepts nature and on providing rules and axioms to help the
identification and grouping of concepts with same nature in a consistent way. The
inspiration behind the idea of identifying concepts nature, which has its roots in
Philosophy, has been used since the 60s in the context of library classification by
Ranganathan, who also adopted fundamental categories to help to identify and group
vocabulary concepts according to their high level nature. In this context, Ranganathans’
categories provide a more intuitive, transparent, although less formal (and consequently
more ambiguous) way to approach categorization.
It is worth noting that if, for one hand, the use of formal axioms can improve
ontology structuring, when applied by ontologists with some expertise in logics and with
some background in Philosophy, on the other hand it can represent a challenge for
63
domain experts to deal with the inherent complexity of such philosophical notions and
the formalisms used to express them [Yu, 2006].
However, although the identification of concepts nature has a major role in
structuring well formed and consistent hierarchies, there is more to it than that. There is
still the need of detailed classification principles to help organize concepts in subclasses,
and, besides, if possible, that those principles could be more easily assimilated by the
community responsible for creating and maintaining the ontologies. In this sense, to the
best of our knowledge, few proposals provide a systematic set of principles to address
the problem. Even so, some are directed specifically to a given subject, such as folk
biological classification of organisms [Berlin, Breedlove, Raven, 1973] or construction
works [ISO DIS 12006-2, 1999], while others [Ekholm, 2002] are focused on
identifying objects properties, which even though helps on identifying facets of interest,
does not present a solution to the issue of organizing them in a more thorough way.
Bodenreider and others (2004) point that there are some principles of good
classification that (biomedical) ontologies are expected to be compliant and that, as they
believe, “rest on a wide consensus among those working on biomedical terminologies”.
Such principles can be summarized as: (i) each hierarchy must have a single root; (ii)
children should have exactly one parent; (iii) non-leaf classes must have at least two
children; (iv) each class must differ from another class in its definition. In particular,
each child must differ from its parent and siblings must differ from one another. The
authors, however, do not present evidences on how long their proposal has been adopted
and how exactly the consensus was reached. Besides, proposing that a hierarchy must
have a single root seems to limit the possibility to express different aspects of a domain,
which, in a different perspective, could be easily presented as facets. Also, their
proposal does not provide guidance to other important aspects of classification, such as
the need to define homogeneous hierarchies, as pointed out by Smith (2005).
As observed, although there has been some concern with the adoption of
systematic classification practices, preventing ad hoc built taxonomies, many of those
practices seem to be recent and still need to mature. Also, some of them lack a more
intuitive and transparent explanation, in order to allow a better understanding by end
users and so, to avoid their rejection.
It is our proposal that Ranganathans’ canons of classification, in use for more
than fifty years, provide a methodological path that can join the convenience of a
comprehensive explanation and a more complete and mature set of guidelines, which
can be easily adopted to help building more consistent classificatory structures. The
usefulness of these canons can be seen in a sample scenario for analyzing Gene
Ontology main classificatory structure, as presented bellow.
3. Ranganathans’ canons
In the present paper, we highlight two sets of Ranganathans’ canons, which provide
guidelines to the organization of classes of concepts: canons for the creation of arrays
and canons for the creation of chains. Chains are vertical series of concepts, which can
be organized hierarchically according to generic-specific relations, or according to part-
64
of relations. Arrays are horizontal series of concepts, organized as siblings in relation to
a parent concept.
In the specific case of the organization of ontologies taxonomical structures, we
have selected a subset of Ranganathans’ canons, of particular relevance to our purposes,
and which we shortly present in the following sections, based on Campos e Gomes
(2008) and also Gomes Motta e Campos (2006).
Some of the canons aim at organizing arrays, as for instance, the canons of
Differentiation, Concomitance and Exclusivity, while others aim at organizing chains,
as, for instance, the canons of Modulation and Subordinate Classes (Ranganathan,
1967a). The canons provide principles that facilitate the creation of classes in a more
consistent way, and, according to Ranganathan (1967a), their violation may result in ill-
formed classificatory structures. The selected canons are explained next.
3.1. Canons for organizing arrays
Ranganathans’ Differentiation canon states that a principle of division used as a
classificatory basis should originate at least two classes. For example, let us consider the
array used to classify catalectic activities of enzymes. That array can have a principle of
division according to the kind of enzymes (hydrolise, isomerase, among others) and
another principle of division according to the kind of reaction catalyzed by the enzyme
(free radical formation, first spliceosomal transesterification, among others).
If the principles of division used to organize the arrays are explicit, it makes the
classification of new concepts easier, as the comprehension of the rationale used to form
the hierarchy helps to figure it out where is the right place for the concept within the
ontology structure:
(...) in a classificatory scheme, concepts that are subordinated to
a more general concept can be grouped more accurately
according to the principle of division that guided this grouping.
Principles of division bring transparency to the vocabulary and
so improve searches, locating and relating the concept according
to its inner characteristics. [Novellino, 1996, p.1].
Ranganathans’ Concomitance canon states that two different principles of
division should not result in the same array. For example, if we adopt the criteria of year
of birth and age to classify a set of individuals, we will have as a result arrays
constituted by the same elements.
Ranganathans’ Exclusivity canon states that elements belonging to an array
should be mutually exclusive, i.e., disjoint in relation to elements belonging to another
array. For example, the term multidrug transporter activity should not be
subordinate to both arrays transmembrane transporter activity and drug
transporter activity. Even if those arrays are organized according to different
principles of division (in the above example, according respectively to the principle of
the local – transmembrane – where the transport occur and according to the kind of
element – drug – which is transported).
65
3.2. Canons for organizing chains
As classificatory principles for chains, we highlight the following canons of
Ranganathan: Modulation and Subordinate Classes [Ranganathan, 1967a] as explained
next.
The Canon of Modulation states that within a hierarchical classificatory
structure of concepts there should be a gradual specificity when organizing concepts in
chains, allowing thus a “conceptual consistence between the classes of concepts”
[Gomes, Motta e Campos, 2006]. For example, let us consider the terms helicase,
ATP-dependent RNA helicase and ATP-dependent DNA helicase. According to
the Canon of Modulation, the last two terms should not be directly subordinated to the
term helicase. It should exist, between the first and the last two terms, a term like ATP-
dependent helicase.
The Canon for Subordinate Classes states that in a hierarchy of classes, the
classes nature should be the same, i.e., they should conform to the perspective adopted
as principle of division that guides the organization of the array. For example, in GO,
the definition of the term binding indicates that the principle of division of its array has
to do with interaction between molecules. This makes us believe that terms like
bacterial binding (Interacting selectively and non-covalently with any part of a
bacterial cell) and “extracellular matrix binding” (Interacting selectively and non-
covalently with a component of the extracellular matrix) fall in conflict with such
principle, for a bacterial, which is an organism, and an extracellular matrix, which is a
cellular component, both have different natures from “interaction between molecules”,
which is a process. According with Gomes, Motta e Campos (2006), this canon
complements the Canon of Modulation, and, if it is not violated, the affiliation sequence
of the chain to the array is correctly assured.
4. Analyzing GO (Molecular Function) and observing Ranganathans’ canons
Our considerations about the utility of Ranganathans’ canons are made in the context of
the analysis of the ontological commitment of the Molecular Function branch of GO.
Ontological commitment can be defined briefly as an agreement shared by a
community about the consensual meaning intended for the ontology, not only
considering its comprehension by humans, but also considering its computational
processing by software agents. We assume that this commitment is not always precisely
explicit, however it can be identified, although partially, by means of the existing
ontology documentation, the analysis of concepts definition and the metadata associated
to ontologies terms. On retrieving the ontological commitment, we expect to have as
results the criteria observed as classificatory principle for the organization of first level
hierarchies of GO’s Molecular Function branch, together with the problems observed as
consequence of the adopted classificatory approach.
The choice of GO´s Molecular Function branch was due to the fact that this
branch has intermediary complexity, if compared to GO´s Cellular Component (less
complex) and Biological Process (more complex), as we could notice when analyzing
the terms definition and also the composition of the first level hierarchies (considering
66
the hierarchies’ depth, number of terms and existing relationships). In this sense, this
choice is adequate to our purposes, for if, on one hand, it provides richness of issues to
explore, on the other hand it minimizes the complexity, already big, of the analysis of
the ontological commitment of GO.
The analysis of the ontological commitment is made upon the analysis of the
ontology’s hierarchical structure, the terms nomenclature, and especially of the terms
definition, which is of main relevance to the formation and comprehension of domains
classificatory structures, once they provide rich semantics about intended meaning of
concepts.
For the sake of space, however, we are not going to reproduce the definition of
subordinate terms, but only data about the first level term being analyzed, and the final
result obtained, i.e., the criteria observed as classificatory principle, along with the
summary of problems found, considering the array being analyzed. In order to help
understanding the analysis of the binding array, we present on Figure 1 its immediate
subordinate classes. The binding array contains more than 1000 subordinate classes.
Figure 1. Subordinate classes of GO´s Molecular Function binding array
Results of the Analysis of the Binding array
Identification: GO:0005488 - Binding
Definition: The selective, non-covalent, often stoichiometric 1, interaction of a molecule
with one or more specific sites on another molecule).
1
The quantitative relation of the products and reactants of a chemical reaction in the proportion they
appear in the chemical equation which describes the reaction [Smith et al., 2000].
67
Narrow Synonym: ligand
Classification2 criteria observed:
• Chemical elements (ex: boron binding), organical compounds (ex: lipid
binding), non organical compounds (ex: nitric oxide binding), kind of ion
(ex: ion binding), organical radicals (ex: acyl binding), clusters of atoms
(ex: metal cluster binding), role of molecules (ex: antigen binding),
cellular locations (ex: cell surface binding).
In order to obtain the classificatory principles, besides analyzing the ontology
hierarchy and terms definition, it was necessary to refer to specialized literature, in order
to understand the nature of certain terms. For example, in the case of the term Acyl
binding (Interacting selectively and non-covalently with an acyl group, any group
formally derived by removal of the hydroxyl group from the acid function of a
carboxylic acid), we came to the conclusion that “acyl” refers to a group of atoms (or
radical), due to the fact that GOs3 term definition refers to the term acyl group, which
can be understood as a group of atoms or radicals (see definition of group and acyl in
Oxford Dictionary of Biochemistry and Molecular Biology) [Smith et al., 2000] 4.
It is worth remembering that the analysis of GOs hierarchical structure, if carried
out by a domain expert, or if applied more thoroughly and deeply in the ontologies
hierarchies, could bring richer results, possibly with a wider range of observed
problems.
Problems observed:
• Violation of the Canon of exclusivity
Example:
o norepinephrine binding: Interacting selectively and non-
covalently with norepinephrine, (3,4-dihydroxyphenyl-2-
aminoethanol), a hormone secreted by the adrenal medulla
and a neurotransmitter in the sympathetic peripheral
nervous system and in some tracts of the CNS.
o This class is subordinate both to alcohol binding (binding/alcohol
binding/norepinephrine binding) and to amine binding
(binding/amine binding/ norepinephrine binding).
• Violation of the Canon of modulation
Example:
o nitric oxide binding: Interacting selectively and non-
covalently with nitric oxide (NO).
2
For each organization principle observed, we have put in parenthesis, and italics, a term exemplifying
those principles.
3
Interacting selectively and non-covalently with an acyl group, any group formally derived by removal of
the hydroxyl group from the acid function of a carboxylic acid.
4
This dictionary appears as bibliographic reference within comments in GOs terms.
68
o Nitric oxide is a drug, according to the definition of the term drug
binding (Interacting selectively and non-covalently with a drug, any
naturally occurring or synthetic substance, other than a nutrient, that,
when administered or applied to an organism, affects the structure or
functioning of the organism; in particular, any such substance used in the
diagnosis, prevention, or treatment of disease) and literature [Gerlach and
Falke, 1995]. Therefore, according to the canon of modulation, the term
nitric oxide binding should be subordinated to drug binding.
• Violation of the Canon of subordinate classes
Example:
o A trisaccharide (Interacting selectively and non-covalently with any
trisaccharide. Trisaccharides are sugars composed of three
monosaccharide units) is an oligosaccharide (Interacting selectively
and non-covalently with any oligosaccharide, a molecule with between
two and (about) 20 monosaccharide residues connected by glycosidic
linkages), therefore its subordination to the class sugar binding
(Interacting selectively and non-covalently with any mono-, di- or
trisaccharide carbohydrate) violates the canon of subordinate classes, i.e.,
trisaccharide binding should be subordinated to the class
oligosaccharide binding.
When tabulating the problems found, we highlight the importance of the canons
of exclusivity, subordinate classes and modulation, as having the greater number of
violation occurrences (7), followed by the canon of differentiation (5). There were not
found evidences of violation of the canon of concomitance (although it is worth
remembering that the analysis conducted did not cover thoroughly the complete
deepness of GO´s hierarchies).
Table 1. Total occurrences found on analyzing GO first level classes
Canon Total violations
Canon of Differentiation 5
Canon of Concomitance 0
Canon of Exclusivity 7
Canon of Modulation 7
Canon of Subordinate Classes 7
The analysis of first level hierarchies of GO´s Molecular Function branch shows
a diversity of problems, which are materialized in a variety of non uniform classificatory
principles observed, which seems to indicate a lack of adoption of well defined
classificatory principles, gap which could be fulfilled by the adoption of Ranganathans’
canons. In particular, we have observed the violation of Ranganathans’ Exclusivity
Canon, which points to the relevance of understanding existing perspectives to think
69
about the nature of the domains concepts. In contrast, we have not observed5 the
violation of the Canon of Concomitance. This finding could be evidence that some
classificatory principles are more intuitively assimilated than others, but could be as
well due to the characteristics of the domain, or yet, due to the deepness of the analysis
conducted.
5. Conclusion
Ontology construction, although a maturing research field, still faces many challenges,
especially in domains with a rich variety of complex concepts, and whose knowledge
advances dynamically. In Biomedicine, for example, the need to organize concepts in a
systematic way has to cope with the pragmatic nature of its community, with no deep
knowledge of Ontology related disciplines such as Logics and Philosophy, but with urge
to improve their ontologies.
One of the challenges of ontology construction is the creation of classificatory
structures, or the backbone taxonomy, with its subclasses organized in a systematic way.
The challenge presents itself not only due to the complex nature of the domains, but also
due to the interdisciplinary nature of ontology building, which demands knowledge of
experts in knowledge organization, such as Computer Scientists and Information
Scientists, but especially, end users who detain the knowledge of the domain and the
intended meaning of concepts contained in their ontologies. Considering that it is
important to provide the grounds for an effective dialogue between people with different
backgrounds and, at the same time, to provide principles that can rapidly and easily be
assimilated and adopted on ontology structuring, it is important to overview existing and
previously successfully adopted initiatives. In this sense, Information Science can
provide relevant contribution, as it has mature hands on experience of information
organization for more than fifty years, classifying subjects from many different areas.
Ranganathans’ canons, some of those (but not all of them) were presented in this
paper, can bring an important contribution to ontology structuring as it provides a
comprehensive set of principles, that can be easily assimilated, and that provide
effective guidance to avoid many of the problems found in ontologies structures, as we
could observe by analyzing Gene Ontology. Although Ranaganathans’ canons have been
originally proposed long ago, they are a mature set of guidance that have been used
successfully by more than fifty years, and are still current and relevant nowadays, as we
have presented on our application scenario.
References
Berlin, B., Breedlove, D. E., & Raven, P. H. (1973) “General principles of classification
and nomenclature in folk biology”, American Anthropologist, n.75, p.214-242.
Bodenreider O, Smith B, Kumar A, Burgun A. (2004) “Investigating subsumption in
DL-based terminologies: a case study in SNOMED CT”, In: Hahn U, editor. KR-
MED 2004; 2004. Whistler, Canada: AMIA; p. 12–20.
5
It is worth noting that with the help of a biologist, with deep knowledge of the subject, other violations
could have been found.
70
Campos, M. L. A. ; Gomes, H. E. (2008) “Taxonomia e Classificação: princípios de
categorização”. Datagramazero (Rio de Janeiro), v. 9, n. 1.
Dahlberg, I. (1978b) “A referent-oriented, analytical concept theory of Interconcept”,
International Classification, v. 5, n. 3, p. 122-151.
Dahlberg, I. (1983) “Conceptual compatibility of ordering systems”, International
Classification, v. 10, n. 2, p.5-8.
Dahlberg, I. (1978a) “Teoria do conceito”, Ciência da Informação, v. 7, n. 2, p. 101-107.
Ekholm A. (2002) “Principles for classification of properties of construction objects”,
Distributing Knowledge in Building - CIB W78 Conference.
Gene Ontology Consortium. (2001) “Creating the gene ontology resource: design and
implementation”, Genome Research, v.11, n.8, p. 1425-1433.
Gerlach, H., Falke, K.J. (1995) “The therapeutic role of nitric oxide in adult respiratory
distress syndrome”, Current Anaesthesia & Critical Care, v. 6, n. 1, p. 10-16.
Gomes, H.E., Motta, D.F., Campos, M.L. (2006) “Revisitando Ranganathan : a
classificação na rede”, www.conexaorio.com/biti/revisitando.htm, Janeiro de 2011.
Guarino, N.; Welty, C., (2004) “An Overview of OntoClean”, In: S. Staab, R. Studer
(eds.), Handbook on Ontologies, Springer Verlag, p. 151-159.
Guarino, N.; Welty, C., (2002a) “Evaluating Ontological Decisions with OntoClean”,
Communications of the ACM, 45(2), p.61-65, 2002a.
Guarino, N.; Welty, C., (2002b) “Identity and Subsumption”, In: R. Green, C. A. Bean,
S. Hyon Myaeng (eds.), The Semantics of Relationships: An Interdisciplinary
Perspective, Kluwer, p.111-126.
Guizzardi, G., (2005) “Ontological Foundations for Structural Conceptual Models”,
Doctorate Thesis in Computer Science, University of Twente, Enschede, Holand,
ISBN 90-75176-81-3, 416 p.
Guizzardi, G.; Wagner, G.; Sinderen, M. (2004) “A Formal Theory of Conceptual
Modeling Universals”, In: Proceedings Of The Workshop On Philosophy And
Informatics (WSPI), Cologne, Germany.
ISO DIS 12006-2 - International Standards Organization. (1999) – “Organization of
information about construction works — Part 2: Framework for classification
of information”. Swiss.
Novellino, M.S.F. (1996) “Instrumentos e metodologia de representação da
informação”, Informação & Informação., Londrina, v.1, n.2, p.37-45.
Ranganathan, S. R. (1967a), Prolegomena to Library Classification, New York : Asia
Publishing House.
Ranganathan, S. R. (1967b) “Hidden roots of classification”, Information Storage and
Retrieval, v. 3, n.4, p. 399-410.
Ranganathan, S.R. (1951), Philosophy of library classification. New Delhi: Ejnar
Munksgaard.
71
Ranganathan, S.R. (1963), Colon Classification. Bombay: Asia Publishing House.
Smith, A.D.; Datta, S.P.; Smith, G.H., Campbell, P.N.; Bentley, R.; Mckenzie, H.A.;
Bender, D.A.; Carozzi, A.J., Goodwin, T.W., Parish, J.H.; Stanford, S.C. (2000),
Oxford Dictionary of Biochemistry and Molecular Biology, Oxford University Press
Inc., New York.
Smith, B. (2005), “The Logic of Biological Classification and the Foundations of
Biomedical Ontology”, In: Westerstahl, D. (ed.) Invited Papers from the 10th
International Conference on Logic, Methodology and Philosophy of Science.
Smith, B.; Kumar, A. (2004) “On Controlled Vocabularies in Bioinformatics: A Case
Study in the Gene Ontology”, BIOSILICO: Drug Discovery Today, v.2, p. 246–252.
Smith, B.; Williams, J.; Schulze-Kremer, S. (2003), “The ontology of gene ontology”,
In: AMIA Annual Symposium Proceedings, p. 609-13.
Yu, A.C. (2006), “Methods in biomedical ontology”, Journal of Biomedical Informatics,
n.39, p.252–66.
72