The Ontology of Descriptions and Observations for Integrated Modelling (ODO-IM) is a core ontology developed to support integrated semantic modeling in k.LAB , a semantic web software platform for science. ODO-IM provides ontological grounding for the scientific models produced by modelers in k.LAB. The ontology is developed and maintained by the Integrated Modelling Partnership in support of its commitment to modular, distributed, semantically explicit, and integrated scientific computing and modeling. The modularity of ODO-IM is achieved through a series of decomposable knowledge-bases that allow for extensibility and reusability according to the FAIR principles - Findable, Accessible, Interoperable, and Reusable.
ODO-IM is an ontology of scientific observations and the scientific descriptions of such observations; the ontology assumes a phenomenological, descriptive, and linguistic perspective and, like other ontologies (e.g. the Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE) ), remains agnostic concerning the metaphysical status of the world in a fundamental sense. In this ontology scientific observations are interpreted and defined as contextual relationships between two entities, an observer and an observation target, which could be a physical object and an event. This type of observation relationship needs to be constrained within the current scientific practices that (very) often employ scientific artefacts, such as datasets and data models, to carry out observations. What ODO-IM postulates is that:
scientific products and explanations result from observations that are based on perspectives, including scales and contexts.
For this reason, ODO-IM is concerned with the definition of Observable
concepts, Predicates that specialize Observables, and the Descriptions
of the scientific processes that produce scientific artifacts. Note that
ODO-IM does not explicitly mention the distinction between universal and
particular that is often integral to foundational ontologies , such as
in the Basic Formal Ontology (BFO) and the Unified Foundational
Ontology (UFO) . While a commitment to universal and particular is not
part of ODO-IM, concrete and abstract categories are implicitly
included; indeed Observables and Predicates abstractions which can be
concretized to compute an observation. For example, when a
scientific description invokes a countable observable, such as an insect
or a plant, the concept is reified. The inclusion of the
concrete-abstract distinction is implicit because there are no
super-classes for concrete or abstract entities in the ontology.
Instead, those emerge when a scientific description is identified and
realized by the system. In this sense, a concrete entity is considered a
token-instance, while its abstraction is analogous to a type . For
example, the Elio is the token-instance of the type CAT.
The three backbone categories of ODO-IM are (i) Description, (ii) Observable, and (iii) Predicate (see Figure #fig:backbone[1] representing the conceptual model the ontology). The ontology builds upon the Provenance Ontology (PROV-O) that aims to trace origins and changes of information and thus are crucial for representing scientific knowledge. (i) Description captures the activities that produce scientific artifacts derived from contextualized observations. Sub-classes of Description can be used to describe the scientific assertions that incarnate Observables into observations within a context, such as models and computational workflows. (ii) Observable are concepts that serve as the object of a scientific Description and provide the basic semantics for the resulting observation. (iii) Predicate can be combined to Observable to create restrictions on the meaning of the Observable.
image general.png[The ODO-IM conceptual backbone]
As anticipated, in ODO-IM, scientific observations are phenomenologically-driven and dependent upon perspectives, contexts, and scales, which means that the observer can decide the focus of the observation, then its description, on structural aspects (i.e. continuant-like object) or functional aspects (i.e. occurrent-like object). For this reason we do not commit to any specific foundational ontology that would bind the whole project to one worldview (for instance notions of continuants and occurrents, or particulars and universals).(Note: See for a general overview of foundational ontologies and their applications.] Each observation can be made under a static (structural) or a dynamic (functional) perspective. This imply that in ODO-IM the viewpoints of the observers and the ``x qua'' Observable have ontological priority over the classification of entities beyond the observation context.
ODO-IM captures the scientific process following the standard provenance
terminology and provides constructs that enable explicit semantics into
the provenance model of PROV-O, building the conceptual foundation for
the semantic web of scientific observations implemented in k.LAB. At the
core, the scientific process is seen as the transformation of existing
knowledge artifacts into others that incorporate and define scientific
advancement; such transformations are enacted by agents as scientific
activities, such as measurement, symbolic integration or numeric
simulation, under the guidance of plans representing the script
for the transformation and incorporating the scientific hypotheses and
methodologies guiding the process. These artifacts, agents, activities,
and descriptions have direct counterparts in the PROV-O concept
hierarchy, namely: entity, agent, activity and plan.
To enhance the provenance model with an explicit observational
grounding, we must recognize that any scientific artifact reflects a
chosen overall entity of the world, providing the spatial and
temporal aspects for successive observations. Therefore the first step
of any scientific activity is the choice of a context of reference,
including its identity (e.g. a model organism, a region of space) and
the required granularity of description for any successive observations
in it, such as the representations of time and space (scale) adopted
for the observation. With this in mind, ODO-IM extends the fundamental
PROV-O classes as follows:
- Resource is the ODO-IM sub-class of PROV-O entity, representing
any communicable, storable artifact that embodies scientific knowledge -
such as a dataset, report, or equation. Resources are always
semantically characterized by the concept they describe (the Observable)
and are intended as directly
actionable, i.e. machine-readable in all their parts and implications with no human intervention. - Contextualization is the ODO-IM sub-class of PROV-O activity, representing the production of a Resource (using zero or more other Resources) and recognizing that a Resource that serves as context must be provided (i.e. acknowledged) by the observing agent to bootstrap the process.
- Description is the ODO sub-class of PROV-O plan, which links a
semantic statement (the Observable) to a computational strategy to build
the Resource mentioned above that represents it - the k.LAB software
stack defines an AI-driven process that finds, connects, and uses
Descriptions from a networked, distributed repository to respond to
logical queries of the form
observe _observable_ in _context_.
Based on these definitions, ODO-IM grounds a conceptual model to facilitate:
. the definition of Observable concepts and their combinations through
semantic operators and Predicates that intuitively capture complex
meanings and linguistic distinctions commonly adopted in day-to-day
scientific practice;
. the definition of scientific methodological recipes (models) as
Descriptions that have Observables as inputs and outputs, representing
the synthetic statements of scientific methods that are typically only
expressed as non-directly actionable artifacts, such as equations in
scientific articles or complex computer code.
The conceptual architecture of ODO-IM is not used directly by practitioners but is instead perused by writing statements in the knowledge-Integrated Modeling language (k.IM) [more information forthcoming). k.IM is supported by the k.LAB software stack and defines ODO-IM compatible axioms using syntactic rules and conventions closely modeled on the English language. The use of k.IM facilitates semantic annotation and makes the building of integrated semantic web applications accessible to the largest possible set of practitioners.
ODO-IM core ontology has been designed to capture scientific
observations encompassing complex epistemological dimensions of
discovery, classification, and conceptualization of entities based on
several activities, such as measurements and evaluations. Scientific
observations are themselves informational artefacts , typically encoded
and stored in physical supports, and represent a particular state of
affairs based on the perspectives, goals, hypothesis, knowledge, and
capabilities of the observer(s). Consequently, observations are
cognitive abstractions guided by empirical and scientific practices
.Yet, the label observation carries an ambiguity that lies between
its meaning in terms of a type of activity (doing an observation)
and a type of information, i.e. a content, that can be replicated,
copied, transcribed, and analyzed to create more content.(Note: An
up-to-date review of information entities in ontologies is presented by
Sanfilippo .] Thus it is important to remark that in k.LAB scientific
observations are formal contextualized information/semantic content and,
in this regard, scientific observations cannot exist without a context
(and one or more observers).
Observables and Predicates are abstract concepts(Note: Here
concept is intended as type rather than universal .]
employed to describe scientific observations that can be computed
according to a Description to produce a Resource. A Description provides
a plan to contextualize an observation into a Resource that represents
it within a spatio-temporal and semantic frame of reference that
provides a context for its computation (the Description that produces
the context itself is a specialized sub-class of Description,
Acknowledgement - i.e. commitment, whose computational aspects are
trivial). The next sections illustrate the main components that
constitute a scientific observation description in k.LAB, namely
Observable, Predicate, and Description and their associated definitions.
The notion of observable has been debated for a long time in several
academic literature. In physics, in which probably the standard
definition of observables was proposed, an observable is a physical
measurable property , such as temperature and mass. In philosophy, the
nature of observables was the object of discussion, in particular by
logical empiricists who attempted to distinguish and define different
aspects of the scientific practice, such as sensory/perceived objective
evidence and unobservable theories . Despite the efforts in defining
observables, especially in the philosophical corpus, what can be
accounted as such has been the object of disagreement. For example, are
observables only entities that can be directly observed through the
perception, or should we include in the catalog of observables also
entities that are observed or measured through the mediation of more or
less complex technologies, such as a thermometer, a water monitoring
buoy, satellite observations, and a magnetic resonance imaging (MRI),
and if yes, how the observations derived from complex technologies are
connected with their unobservable preconditions based on articulated
theories ? Recently it has been suggested that perhaps the idea of
observations conceived in the old-fashioned way is not any-more
appropriate and useful . Leaving aside (i) the strict direct versus
instrumentation-mediated observability issue, which it has been
partially surpassed also due to the contemporary over-reliance of
technology in most of the scientific fields, and (ii) the philosophical
literature that argues in favor of a broader and often pervasive role of
technology in our lives, such as for postphenomenology , a more modern
reading of observables is through empirical/observation data , and
to extend observations from direct empiricism to a more complex
interpretations .
The relation between data and phenomena has been discussed in a
seminal work written by Bogen and Woodward , in which data is described
as an observable evidence for the phenomena - although data cannot be
explained by theories - the phenomena is instead rather unobservable and
accessible through data, yet theories explain phenomena and not data
(pp. 305-306). This view has been re-elaborated by Votsis , who
considers data as evidence for theories when auxiliary hypotheses are
adequately employed. In a clarification of the work of Bogen and
Woodward, Teller offered a contemporary analysis of the phenomena
following the hierarchical structure of models proposed by Giere in
which the relations between world, data, theories, and models are
articulated. In particular relevant for our discussion is the idea that
physical quantifiable qualities that are included in models(Note: In
this writing we will not discuss the different kinds of models presented
in Giere .), such as mass, temperature, and weight, are (i) interpreted
and contextualized within the scientific enterprise, and (ii)
related to real-world entities to provide the basis for further
interpretations .
In the context of k.LAB the dichotomy between data and phenomena, and
observable and unobservable is somewhat unnecessary since scientific
observations are, as already mentioned, information entities derived and
elaborated through hypothesis and contextualization of other scientific
artefacts (e.g. datasets, data models, and images) taken as
phenomenological evidence and interpreted based on scientific
humanly-driven perspectives , perspectives that are also domain-based .
Thus ODO-IM Observables, with the capital O that refers to the concept
observable, are central elements for describing observation and are
defined as abstract concepts that represent entities that can be
observed through the mediation of scientific artifacts, such as physical
objects, processes, and qualities. Examples of Observables are city
and land cover type, which are classified according to ODO-IM as
subject and quality. Note that in ODO-IM Observables are agreed
upon and negotiated within scientific communities and should be also
negotiated beyond them in a transdisciplinary effort. In the following a
description of the main Observable's concepts is provided, more
specifically: (i) countable, (ii) process, (iii) quality, and
configuration (see Figure #fig:observable[2]).
image observables.png[ODO-IM taxonomy of Observables]
In philosophy, countable entities are often called
sortals.(Note: See the Sortal for a richer and articulated
review .] countable can be identified as individuals carrying a
principle of identity and having unity, then countable are wholes
having boundaries , i.e. temporal/spatial/conceptual. So when we refer
to countable, we answer the question How many x are there?, for example, how many people and lakes are in a certain
region? In k.LAB countables are reified by instantiation Description
and are organized into: (i) substantial, (ii) event, and (iii)
relationship.
- A substantial (here adopted in a similar fashion of UFO and DOLCE ) is
a concept that captures endurant-continuant like entities, such as
lamb,mountain, andperson. Substantials, which are often understood with the less technical term object, describe entities that endure in time by being wholly existent and bear qualities. In k.LAB only substantials can determine the context for observations; indeed the first Resource in any scientific process modeled according to ODO-IM must be acknowledged (i.e. explicitly declared). Thus, despite the assumed observational perspective (structural-functional), this aspect manifests the overall ontological priority of substantials over non-substantials, such as events and processes. Subject and agent are sub-classes of substantial, while the latter carries agentive characteristics, i.e. intentionality , the former does not. - In the ontological literature, events are typically classified as
perdurants/occurrents (e.g. ), which corresponds to event instances,
i.e. tokens, that happen in time, are composed of temporal parts, and
involve participants . However, in the case of ODO-IM, events are
Observables inheriting their abstract-type
repeatableorreusablecharacteristics. The notion of repeatability has been analyzed by Galton in the context of processes as abstract patterns , ODO-IM adopts a similar view for events in which those are cognitively salient abstract scripts that can be reused to described and give structure to one or more qualitative change in substantial Observables.(Note: Although each of these accounts of events present analyses relevant for ODO-IM, nevertheless, these works are not entirely adherent with events described in ODO-IM. In particular, in the recent paper of Guarino et al., , the examination of events from a metaphysical and semantic perspective is way more detailed and refined than the one provided in ODO-IM in which, for example, at this stage of development, interpretations of modifiers are not provided, and the description of the individual (token) qualitative changes is not relevant for our current discussion.] The idea of events as abstract scripts is inspired by the influential paper of Schank and Abelson , in which scripts are descriptions of a series of contextualized events where participants play roles. Many of the scripts considered in refer to planned intentional events; however in the context of current k.LAB applications, events are also natural and thus might be generated outside human interventions. ODO-IM events are identified by (i) their temporal boundaries, a start and end point (over a time interval ) which limit the time-span of the event in a cognitive relevant interval, (ii) the presence of substantials that participate to the events and exhibit changes, and (iii) as being made by processes (this aspect of event has also been called durative ), for example, the eventseasonis made by processes such asprecipitation. The relation between events and processes opens the debate concerning the level of description of those concepts; this is discussed below when we introduce the process construct. Examples of events arebirth,conference, andwildfire; considering the latter (i.e.wildfire), the qualitative change is described, for example, as the water vaporized from the vegetation. - A relationship(Note: In this writing, we do not engage in the
discussion concerning the difference between
relationandrelationship.] is a countable concept directionally connecting two substantials. Relationships in ODO-IM are structural and functional whether the observer’s focus is on static or dynamic aspects of the relationship. Structural relationships between two substantials (relata) allow another substantial (relator) to emerge. Note that the connection between two relata is what allowed the relator to emerge, i.e. to be observed, an interpretation that is inspired, although with some significant differences from Guarino and Guizzardi .(Note: Among the most prominent differences, in the work of Guarino and Guizzardi , the connection between the relata (for the author therelation) is derived from the relator that is what bound the relata being its truthmaker. Instead in ODO-IM the relata is dependent upon the relationship itself.] For example,city A is connected to city Bgenerates the subjectroadthat links the two cities. Focusing on functional relationships, those generate events, for instance, the functional relationshipskier using a mountain (for recreation)comprises interaction(s) between the subject skier and mountain, which generates for instance the eventski slalom. So structural relationships engender or define subjects (e.g. parenthood between human individuals engenders families) qualities bearer, and functional relationships engender/define events that are composed of processes.
As mentioned before, events and processes are connected as the former is
composed of the latter. One of their linkage examined in the ontological
literature that is pertinent to ODO-IM is the perspectival nuances that
involve the events and processes descriptions . While events are
countable boundary objects, processes are described as open-ended and
homogeneous . In addition, processes are described as experienced
dynamically at a certain time, while events are historical records that
can be remembered . Both event and process are involved in changes, for
example, the former triggers something else, such as another event, and
changes occur to its participants (change of state ), the latter
perpetrates those changes in a state of change mode. Thus
while storm and earthquakes are events with a start and
termination points, pollination is an ongoing course that is
described with a focus on its on`going interactions or affects. Processes
in k.LAB are realized through simulation Description.
Observable qualities require an intermediate entity to produce a
Resource that represents it, e.g. a reference is needed to carry out
a measurement. In this regards, the intermediate entity, for example a
substantial, bears quality, or from the other way around, qualities are
inherent to the intermediate entity . For example temperature cannot
be measured without a reference entity, such as water, atmosphere and a
substantial body. In addition temperature can be measured only in
reference to a conventional unit or other numeric scale in order to
produce a quantifiable outcome, such as 25 degrees Celsius.
The hierarchy of quality is organized following the sub-classes of (i) contextual, (ii) presence, (iii) enumerable, and (iv) quantifiable. (i) Contextual quality defines spatial and temporal aspects of a description providing its context. Both are required to frame the scale in which a description is given. From a phenomenological point of view, the experience of time involves factors that range from duration to granularity, ordering, and change ; nevertheless, ODO-IM focuses on the informational representation of time as a quality that defines the granularity and extent of a description. Following this target, in this ontology, time is treated as a way to capture temporal features such as the duration of an event and change over a specified interval of reference, for example, months or years . Spatial qualities are also accounted for with a similar topological description, noting important differences compared with temporal qualities, for example, variable dimensionality. The notions of extent and granularity, as well as operations such as union and intersection, are similar in both dimensions and can be accounted similarly. In ODO-IM spatial characteristics are represented in terms of regions of space that can be associated with intervals to create temporal and spatial notions. In addition to space and time, other dimensions may be used to account for conceptual topologies (e.g. a multi-hypothesis space) over which observed states. Consequently the representation of Resources may be distributed.
While (ii) presence specifies the existence of the quality itself, a semantic equivalent of a boolean “truth value", (iii) enumerable quality defines the quality on based on a resemblance with others of its kind, thus providing quality classification. Differently from the other qualities, (iv) allows for the expression of more fine-grained measurable qualities by referring to a unit of reference or fractions and therefore is, assigned to a numeric value when an observation is concretized, e.g. volume and mass. Thus quantifiable qualities refer to all those qualities that can be quantified and are contextualized by quantification Descriptions and includes several sub-classes ranging from continuous numerically quantifiable quality, which includes important notions such as physical property and probability, to value that specifies monetary and preference qualities referencing a specific value attribution and trading system (see e.g. ). Continuous numerically quantifiable quality are divided into: physical property, probability, relative quantity, and uncertainty.
- Physical property is contextualized by measurement Description and classifies qualities of physical entities, such as substantials. Physical properties are divided into extensive and intensive; while extensive physical properties are influenced by the physical structure of the inhered entity, the same does not hold for intensive physical properties.(Note: For more details see the IUPAC Compendium of Chemical Terminology and https://en.wikipedia.org/wiki/Intensive_and_extensive_properties.] Examples of the former are volume and length, instead of the latter temperature and duration.
- Probability is the measurement of the likelihood of occurrence of
favorable eventsthat is contextualized as a probability description. Probability has one sub-class, called presence probability, that frames the probability focus on subjects in a certain context. - When a quality requires another (compatible) quality to be assessed is
called relative and has a similar semantics as
integralqualities in the conceptual spaces theory, which mentions the example of hue and brightness when considering the color of an object . Ratio (contextualized by ratio description) and proportion are sub-classes of relative quantity. - The last continuous numerically quantifiable quality is uncertainty; as mentioned by Floridi , the scientific literature provides several definitions of uncertainty, such as in physics, mathematics, statistics and epistemology, yet generally speaking uncertainty can be defined as partial epistemological content (i.e. information) concerning a state or a result . In ODO-IM, uncertainty is contextualized by uncertainty description and is treated as a quantifiable quality that can be assigned to a description, with no further assumption about the method of measurement or range of values.
Besides continuous numerically quantifiable, other sub-classes of quantifiable qualities are numerosity, priority, and value.
- Numerosity defines the number of countable things in a group and is contextualized by the counting Description.
- Priority describes a monotonic ordering of concepts that is contextualized in terms of a numeric ranking Description.
- Finally value(Note: We acknowledge the abundant and complex literature on values, particularly the philosophical one, see e.g. . Yet, in ODO-IM, the value concept has a less ambitious purpose defined within its quantification scope.] is a quality assigned to an entity by a subject, typically an agent, based on specific criteria. Value in contextualized by valuation Description. In ODO-IM values can be either monetary or non-monetary (expressing on a preference) following the ecosystem services classification of values . The assignment of value is expressed as a priority, i.e. abstract ranking, based on the characteristics of the valued entity and can potentially change over time .(Note: The paper of Porello and Guizzardi treats preferences as ternary relations, here we consider unary predicates as the agent making the valuation is always explicitly known in k.LAB and contextual to any observation made.]
The last Observable in this list is configuration, which presents
conceptual and operational differences compared with other ODO-IM
concepts. Indeed configurations are somewhat external to the user’s
control since these constructs cannot be directly introduced while
modeling and can only be created by the system through a detection
Description. Configurations represent emergent patterns that are
generated by observations of quality and relationship concepts.
Configurations are dependent upon the observer that decides the focus on
particular aggregations. For example social network is a
configuration that emerges from the functional relationship ``social
connection''.
As mentioned in Section link:#intro[1), ODO-IM undertakes a
phenomenological, descriptive, and linguistic ontological position of
scientific observations. In particular, scientific observations depend
upon perspectives, more precisely structural and functional, and
Observable concepts live in context, which signifies that
Observables are observed not in isolation but in relation to each other,
allowing the identification of further ontological constraints. Once the
focus shifts on the relationships between Observables both
(a) patterns of dependence and (b)
presence of arity. Both a and b are
connected; while the former specifies whether the Observable is
dependent upon another (based on the observation perspective), the
latter defines absence or presence of dependence and, in case of
presence, its numerosity. Although the ontological literature includes
several other meta-ontological properties , here we focus only on
dependence (and its arity) as it is the most pertinent due to the
observational purpose of ODO-IM.
More specifically, (i) substantials and events are the only Observables that can be considered as guiding perspectives for an observation, either structural if the focus is on the substantials as the main context or functional when an event is the primary context of the observation. In this sense, when an observation is described, substantials and events are independent bounded observables that can be assumed as guiding perspectives for an observation, either structural or functional, respectively. Note that when the observer decides to perform a structural observation, the context is bound to a substantial one, and events can become dependent upon substantial participants; the opposite situation is instead manifested when events are selected as the primary context.
While substantials and events have an arity of dependence for their observability equal to 0 in their corresponding observational perspectives, (ii) qualities and processes always inhere and are dependent upon other Observables, meaning from any standpoint, exhibiting an arity of dependence \geq 1. Indeed while qualities inhere to substantial, processes do to events. Finally, (iii) structural and functional relationships are dependent from both their target and source (i.e. = 2) to be observed and conceptualized as countable. Table #table:table1[1] summarizes the aforementioned points.
| Observation perspective | |||
|---|---|---|---|
| Arity of dependence | Dependence | Structural | Functional |
0 |
No | Subject | Event |
1 |
Yes | Quality | Process |
2 |
Yes | Structural relationship | Functional relationship |
In ODO-IM, Predicates are abstract concepts that (i) cannot be directly
supported by scientific observations and (ii) provide a linguistic tool
to characterize Observables further. For example, important uses for
Predicates are to describe more specific qualities that often cannot be
observed, only attributed, e.g. color in substitution of ``reflected
wavelength'' of an object, and to define arbitrary discrete scales, such
as very short, `short`, `tall`, and `very tall`.
Broadly speaking, predicates are often considered the linguistic
counterpart of properties identifiable using the copula is in
sentences as x is P . This association between the predicate and the
is generated semantic vaguenesses, summarised by the so-called
Frege-Russell ambiguity thesis citing Villko and Hintikka , in which
is carries multiple meanings, namely predication, identity,
existence, and subsumption. Long ago, this issue has also been the
object of debate in knowledge representation and formal ontologies;
notable are the works of Brachman in semantic networks , and Guarino and
Welty on identity and subsumption . ODO-IM mitigates some of the
ambiguities mentioned above by recognizing the difference between
Predicates that describe general aspects of
the Observable, their physical location, their
contextual involvements, and finally some Predicates
categorize epistemic dimensions of the
Observable. (Note: the organization of Predicates in ODO-IM
resolve mostly the ambiguity between general predication and identity.]
Predicate category includes (i) attribute, (ii) realm, and (iii)
role, the latter (iv) domain and (v) identity (see Figure
#fig:predicate[3]).
- Attribute is the most generic Predicate that can be used to specialize
Observables. Attributes describe inference to Observable in the
following ways: (i) as capability or disposition of the Observable, such
as
reproductive,human visible, andpervious, (ii) as states or phases of Observables that have been reached or are ongoing, such aspollinated,irrigated,adult, and finally (iv) as subjective ordering by defining non-quantitative rankings to be attributed to Observables, such assevere,mild,hot,damaged. - Realm restricts the physical location of an Observable, for example,
the temperature of the
atmosphereor the amount of soil in thesoil stratum. This predicate can be intended, in a similar fashion to that of realms included in the International Union for Conservation of Nature (IUCN) , as parts of the biosphere which are divided into terrestrial, subterranean, atmospheric, marine, and freshwater. - Role is a dependent and contingent Predicate that is played by other
entities and is bound to a context. The same entity can play multiple
roles simultaneously in different contexts without affecting its
identity or core semantics. Note that role is different from other
predicates because it describes external characterizations of
Observables (contexts), for example, an insect may be a
pollinatorin one particular ecosystem but not in another.
While the Predicates mentioned above specify either traits that are
contextually or non-contextually ascribed to Observables or the location
where Observables are placed, epistemic predicates have the function
of explaining Observables either based on (i) their domain of
application or (ii) categorization established within scientific
communities following some of the accounts proposed by the so-called
philosophy of science approach .(Note: Here we extend the idea of
natural kinds and their explanatory and inductive roles in science
to other kinds or classifications, such as for artifacts and domains of
knowledge . Note that the explanatory and inductive relevance of
natural kinds is central to the positions assumed by Boyd , which is
far more articulated than our exposition of epistemic Predicates in
ODO-IM [#f1]#[f1]#.] Domain Predicates assign Observables to a
scientific discipline using domain-specific ontologies as reference; for
instance, the Observable quality elevation belongs to the
geography ontology. Focusing now on the most challenging Predicate,
i.e. identity, this concept has been widely discussed in ontology and
metaphysics , and it is outside the scope of this technical report to
address its vast and diverse literature;(Note: In the context of
applied ontology and conceptual modeling, Guizzardi and colleagues
proposed an analysis of the so-called Powertypes, i.e. types having
instances that present both types- and instances-like aspects.
Powertypes exhibit similarities with ODO-IM identities that should be
further investigated in the future.] thus, for this analysis, the aim is
to propose a description of identity as meant in ODO-IM and for its use
within k.LAB software. Identity in this ontology represents abstract
classes of entities with unique identifiers that are grouped (i.e.
classified) based on cognitive relevant aspects, normally subsuming a
large number of characteristics, such as morphological, functional, and
behavioral, for example, concepts that belong to the kingdoms of
animalia and plant, the classification of rocks, and artifacts. To give
a more explicit example, the concepts belonging to the taxonomy of
marine mammals, e.g. cetaceans and pinnipeds, are identities in the
ODO-IM sense. In k.LAB, identities play an important role and are
managed through what in the system is called authority , following a
strategy to link concepts with referenced external terminologies (e.g.
IUPAC for chemistry ), thesauri (e.g. the FAO AGROVOC ) and information
providers (e.g. the Global Biodiversity Information Facility (GBIF) ).
An example is the identity water that associates the meaning of
water in reference to the IUPAC classification .
An interesting final consideration concerning epistemic Predicates is their recursive aspect; indeed, both domain and identity represent categorization systems in a taxonomical fashion that, in a simplified way, correspond to simplified ontologies within the ontology ODO-IM. In other words, epistemic Predicates are themselves kinds of ontologies, more specifically taxonomies. In addition, as mentioned in footnote link:#f1[[f1]), epistemic Predicates have explaining and inferring functions. In particular, the former, i.e. explainability, is a characteristic that recently Guizzardi and Guarino have been attributing to the role of ontologies and ontological analysis for domain models, including in the context of scientific practical explanations (e.g. goal-based explanations), an aspect that should be further analyzed also for epistemic Predicates.
image predicate.png[ODO-IM taxonomy of predicates]
The main concrete element of ODO-IM, except for Resource that is derived
from the PROV-O , is Description (see Figure #fig:description[4]). It
defines one or more Observable aspects within a scale of observation
through a recipe that creates a Resource that represents it in the
selected spatio-temporal context. In ODO-IM Description is a plan
(PROV-O entity), the latter denotes a set of activities performed by an
agent for a purpose . Although often in literature plans have been
defined as types of descriptions , ODO-IM descriptions are specific
plans, specialized according to the Observable, that define a set of
ordered activities undertaken by the system to produce an outcome, i.e.
Resource. More specifically, Descriptions are derived from the
Resource-in-input and Resource-in-output processes.
As outlined above, Descriptions interface with Observables by providing a strategy to contextualize them, thereby setting the conditions for Observables to be observed, thus to exist . Observables do not have temporal or spatial characteristics, yet Descriptions facilitate the concretization in context of Observables when invoked. The category Description is divided into three first-level sub-classes: (i) instantiation, (ii) resolution, and (iii) detection.
- Instantiation. This Description reifies concepts, specifically countable, by building the Resources that incorporate them into concrete token. In this way, instantiation creates concrete concepts. The sub-classes of instantiation are acknowledgment and connection; the former consists of the intention/commitment to generate the concept and directly specifies all details of the outcome. The first step of any scientific process is acknowledging one instance that serves as context for the remaining activities. The latter, connection, contextualizes relationships.
- Resolution. This type of Description explains the concept through its process realization based on a scale: while instantiation only produces a concept at time t by invoking it, resolution initiates a modeling process that results in a corresponding outcome.
- Detection. This class contextualizes a configuration and can produce a model resolution.
Resolution, which describes the explanation of a previously instantiated
observable, is further articulated as simulation and state
attribution; the former is applied in the context of a process, the
latter is a concept container that includes all Descriptions that
attribute values when applied to: classification, quantification
and verification. These three sub-classes of state attribution apply
respectively to (a) enumerable qualities (whose attributed values must
be concepts subsuming a specified conceptual space, e.g. soil types),
(b) quantifiable qualities whose values are numeric, and (c) presence
that can be asserted by means of simple truth values (true/false).
image description_final.png[ODO-IM taxonomy of descriptions]
Quantification is the most articulated, presenting the following subsumed elements(not included in Figure #fig:description[4] for reasons of space): (i) counting, (ii) measurement (iii) probability description (iv) ranking, (v) ratio description that expresses a relationship between two or more numerical quantities that produce another number as an outcome, (vi) uncertainty description which uses numerical data to express the uncertainty concerning another state, and finally (vii) valuation that produces a numeric estimate of the "worth" of an entity in a given context and perspective.
An extensive description of the PROV-O elements can be found in , in
this work we simply report the PROV classes that we included, which are:
activity, entity, agent, instantaneous event and role (and
their sub-classes). Two remarks on entity (Note: the
PROV entity is conceived as a synonym of artefact/informational
object.] are necessary. First we added Description as a sub-class of
plan that is a sub-class of entity, and second we defined another
sub-class of entity that is Resource. In particular, Resource, as
mentioned in the introduction, this element in ODO-IM is generated by
some contextualization of Observables and is any scientific artifact
that lives in the resource layer of k.LAB. Those include literals,
datasets, data services, computations, and computational services
without a semantic characterization. Resources are uniquely identified
by a URN, which are resolved to their contextualized values through
network services.
In ODO-IM while 20 object properties are imported from the PROV-O, for
example prov:wasInfluencedBy that defines the agent-to-agent
responsibility, 21 object properties have been newly created for this
application, those are:
affects: a process affects a quality in its context when it is able to change the state of the correspondent quality as time moves on. If this relationship exists, the state that describes the quality is dynamic.appliesTo: allows restricting the range of observables that an observable applies to.confers: specifies the attribute or role conferred by a process.contextualizes: defines the relationship that links an observation to its observable.creates: specifies the qualities or countable generated by the the process.describesQuality: defines the quality using another quality or an attribute.emergesFrom: establishes observables emerged from quality or countable.increasesWith: defines the increase proportionaliy of quality and ordering..decreasesWith: defines the decrease proportionaliy of quality and ordering.exposesTrait: has as a domain an enumerable quality and as a range a trait.hasDestinationandhasSource: define the links between types of relationship and the class of subjects.hasRole: points to the roles of an observable in the current context.hasScale: specifies the scale of a description.impliesDestination: this object property is used for a role that applies to a relationship, to restrict the type of destination the relationship can lead into.impliesObservable: this property embodies the notion ofimplicationof another observable, whose existence is implied by observing a particular trait. For example, the traitwarmimplies presence of a process in which energy moves particles causing heat, or more prosaically, observability of buildings in a point implies that a building is there.inherits: specifies the predicate inherited in an observable.limitedBy: used to limit the possible value of a categorical observer to a partial union of sub-types.participates: defines the participation of substantial in event or process.relatedTo: most generic stated restriction used to imply an allowed dependency.representedBy: can be used to restrict an observable when it is observed through another - e.g. geographical direction by presence of moss on trees.requires: used for compulsory declared elements to be coupled to an observable.