An observational variable encodes what was measured, observed, derived, or computed in relation to Earth systems and phenomena representation in general. Well defined variables make data easier to find and reuse. However, increasing semantic interoperability of a variable associated concept is still a challenge. In order to avoid inconsistencies and ambiguities between different variable interpretations, it is essential to use a common terminology to homogeneously represent the core elements usually hidden in the variable description or naming. The conceptualization for measurements according to an ontology pattern language (OPL). It establishes standards for representing common core measurement concepts across various application domains. This paper discusses the use of M-OPL in the ontology of the I-ADOPT framework, promoting its semantic enrichment. As a result, we present an I-ADOPT alignment to the patterns established by M-OPL, with additional extension proposals to contemplate the particularities of the measurement domain.
Patterns are instruments for encapsulating common knowledge. The term “pattern language” in the
Software Engineering community refers to a network of interrelated patterns together with a process
for systematically solving coarse-grained software development problems [
An OP describes a particular recurring
modeling problem that arises in specific ontology
development contexts, presenting a well-proven solution for the problem [
2022 Copyright for this paper by its authors.
The Measurement Ontology Pattern Language (M-OPL) addresses the main conceptualization
associated with measurements in general, taking the Unified Foundational Ontology (UFO) as a basis
[
Measurement can be defined as a set of actions aiming to characterize an entity by attributing values
to its properties [
In order to avoid inconsistencies and ambiguities between different domains, it is important to use a
common terminology to represent core concepts shared by them. Ontologies have been recognized as
an important instrument for making knowledge clearer, promoting a common understanding, and
avoiding inconsistencies and ambiguities between different domains. Through a shared
conceptualization, ontologies can play the role of a “contract” established between parties for the
purposes of communication and semantic interoperability [
A core ontology provides a precise definition of the structural knowledge in a specific domain that
spans several application domains in that field [
There are different meanings associated with the term “variable”, depending on the context where
it occurs [
The InteroperAble Descriptions of Observable Property Terminology Working Group (I-ADOPT
WG)3 was responsible for creating a community-agreed framework for representing different aspects
of observable variables like those in environmental research. The I-ADOPT framework ontology was
designed to facilitate interoperability between existing variable description schemes (including
domainspecific ontologies, semantic models and structured controlled vocabularies) [
In this work, we describe the use of M-OPL in the I-ADOPT framework ontology, first identifying concepts and relations that are semantically overloaded in the ontology. The goal is to semantically enrich the ontology, using M-OPL as a reference, to clarify core measurement concepts common in several application domains. By aligning the I-ADOPT ontology to the core modeling patterns proposed for measurements, we aim to capture the conceptualization of measurements for the compound concepts variables. Moreover, this alignment contributes to representing concepts not yet addressed by the ontology, such as scales and units, measurement procedures, measurement planning and measurement analysis.
This paper is organized as follows: section 2 presents the background of the paper, including a review of the I-ADOPT framework ontology and M-OPL. Section 3 describes the use of M-OPL, initially applying five pattern groups in the I-ADOPT ontology to clarify the conceptualization of measurements for compound concepts variables. Additionally, this section presents the relation between I-ADOPT ontology concepts, M-OPL patterns, and UFO fragments. Section 4 describes how M-OPL can be used to extend the I-ADOPT ontology concepts and relations. Finally, in section 5, we conclude and list some future work.
3 https://www.rd-alliance.org/groups/interoperable-descriptions-observable-property-terminology-wg-i-adopt-wg
In this section, we present an overview of the I-ADOPT framework ontology. The I-ADOPT WG was created in 2019 under the umbrella of the Research Data Alliance (RDA) Vocabulary Services Interest Group (VSSIG). In the meantime, we also present an overview of M-OPL pattern groups covering six measurement aspects besides a process suggesting an order to apply them. 2.1.
The I-ADOPT WG had a strong focus on variables observed in environmental research as it
leveraged existing efforts to accurately encode what was measured, observed, derived, or computed in
relation to Earth systems [
The development of the I-ADOPT framework followed a bottom-up approach through phases. The
initial phase was dedicated to the collection of user stories from the environmental domain, identifying
key requirements, and analysing existing semantic presentations of scientific variables and
terminologies in use. The proposed framework was then tested against a variety of examples to ensure
that it could be used as a sound basis for the creation of new variable names as needed. Finally, the
results were formalised into the I-ADOPT ontology and subsequently extended in an output of
recommendations guideline [
The I-ADOPT ontology was inspired by the atomization approach of the “Complex Property
Model” [
Figure 1 shows the schema and instance levels overview of the proposed I-ADOPT framework
ontology [
To explain the concepts and relations at the instance level, we use an example of a complex
biodiversity compound concept variable, adapted from [
As aforementioned, a Variable is a description of something observed or derived as a compound concept, consisting of at least one entity (the ObjectOfInterest) and its Property. The Property (concentration) is a type of characteristic of the ObjectOfInterest. The Entity is an object or occurrence that has a role in an observation. An Entity may play one of the following roles: ObjectOfInterest (endosulfan sulfate), ContextObject (ostrea edulis), or Matrix (flesh). The Constraint (wet) limits the scope of the observation and restricts the context to a particular state. It describes relevant properties of the involved entities in the particular observation. These concepts are interconnected using the following object properties: • hasProperty: It relates a Property with a Variable, with a cardinality of 1..1. This cardinality indicates that the Variable has exactly one Property; • hasObjectOfInterest: It associates the Variable with the ObjectOfInterest, i.e., the Entity whose property is observed. Similar to the previous one, its cardinality is 1..1, meaning that a Variable requires exactly one ObjectOfInterest; • hasContextObject: It associates the Variable with entities that provide additional information regarding the ObjectOfInterest, i.e., ContextObject entities. Its cardinality is 0..*, which means that a Variable may have more than one Entity associated in this context or none; • hasMatrix: It associates the Variable with the Matrix in which the ObjectOfInterest is contained. It is not mandatory, and when it exists, it should only show one Matrix, so its cardinality is 0..1; • hasContraint: It relates to the constraints associated with a Variable, being optional. Its cardinality is 0..*; • constrains: It associates a Constraint with an Entity of the Variable. A Constraint can constrain one or more Entities. Its cardinality is 1..*.
It is important to highlight that in the I-ADOPT ontology, entities assume a role by means of the relation they are associated with. Consequently, the same entity can appear as ObjectOfInterest, ContextObject, or Matrix. Therefore, it could also have different roles depending on the particular variable. Besides, the ontology does not yet cover any additional concept or relation associated with units, instruments, methods, time-related and geographical location information. Units are essential information for describing measures, but a quantitative variable might be expressed differently, requiring units to be modeled independently of variables. Although these concepts provide essential information for interpreting actual observations, they were not originally intended to be included in the scope of the I-ADOPT framework. For this reason, in Figure 1, the context information of “per unit weight” could not be associated with any concept or relation at the schema level of the ontology. 2.2.
The M-OPL was developed following a pattern-oriented design approach. It addresses the core
conceptualization for measurements, according to UFO, and consists of six modules covering the
following aspects [
Following the OPL approach [
MEnt and TMElement patterns are defined to address aspects concerning the definition of measurable entities and properties. The former handles entities and their measurable elements identification. The second pattern defines the measurable element type, i.e., whether the element is directly or indirectly measurable. Directly measurable elements do not depend on others to be measured, such as the body weight. Indirectly measurable elements, on the other hand, depend on other measurable elements, such as the velocity of a body, which depends on the distance traveled and time.
MScale, TMScale, MUnit and MUnit&Scale patterns deal with aspects of scales and units of measurement associated with measures. MScale defines the scale and the constituent values. TMScale pattern establishes the scale types, while MUnit pattern defines the associated Measurement Unit. Finally, MUnit&Scale represents the relationship between units and scales.
MProc, TMProc, MProcBM, MForm, MProcDM patterns address aspects concerning measurement procedures. These procedures describe the steps to collect data for the measures. TMProc is applicable in ontologies that address different types of measurement procedures. MProcBM and MProcDM are patterns for defining procedures for base measures and derived measures respectively. MForm is employed with MProcDM, dealing with measurement formulas to calculate derived measurements.
TMGoal, INeed, MPI, MPI-MP and Ind patterns treat aspects related to measurement planning. TMGoal defines measurement goal and its subdivisions, comprising simple or composed measurement goals. INeed pattern deals with the necessary information identified from the measurement goals. MPI pattern handles measurement planning, specifying which measurement should be collected for each goal or required information. MPI-MP pattern is responsible for choosing the measurement procedure to be used. Ind pattern establishes the measurement that works as an indicator to evaluate the achievement of some measurement goals.
Meas and MAna patterns are concerned with modeling aspects related to Measurement and Analysis. The Meas pattern address the data collection, defining the Measurement concept and its Relations. The MAna pattern is responsible for the analysis.
Figure 2 presents the process proposed in [
In the literature, we found related works using M-OPL in different domains. In [
Here, we present a new domain for using M-OPL, the domain of observational variables. Because these variables, observed in the biodiversity field, require a complex semantic representation to describe the data acquisition and prevent ambiguous descriptions. Like many other domains, biodiversity research has been transformed by a big data revolution, where providing information interoperation is urgently required to support responses towards a sustainable future.
In this section, we present how M-OPL can contribute to clarify and capture the conceptualization of measurements for compound concepts variables in the I-ADOPT ontology, and also more specific aspects concerning measurement in high maturity representation levels. M-OPL ontology patterns were developed taking UFO as a basis. Hence, we also present how the I-ADOPT ontology concepts and relations can be aligned to M-OPL patterns and UFO fragments.
The competency questions presented in section 2.2 were used, as they play a prominent role in defining the scope and purpose of the domain conceptualization, serving as a testbed for ontology evaluation. We also used the same concept example, obtained from NVS, to exemplify the instances addressed to M-OPL patterns.
In order to use M-OPL, we initially selected five pattern groups (MEnt, TElem, Mea, TMea, and MUnit) to be applied, based on the sequence suggested by the process presented in Figure 2. Figure 3 shows a fragment of the application of those M-OPL pattern groups as a proposal to capture the conceptualization of measurements for compound concepts variables. The M-OPL patterns are depicted in gray, followed by the instances of the concept in white. The concept example is illustrated in orange, and it lines emphasize the I-ADOPT ontology relations.
The first pattern applied was MEnt (Measurable Entity), that has been instantiated in order to consider the Measurable Entity for the compound concept variable, answering competency question Q1: wet flesh. Measurable entities are classified according to Measurable Entity Types, that also have been instantiated in order to consider the type of measurable entity relevant to the Component of oyster anatomy domain, answering competency question Q2.
After applying the pattern MEnt, two patterns were applied in parallel: TMElem (Types of Measurable Elements) and Mea (Measures). Through the pattern TMElem, we could identify the Measurable Element for the compound concept variable, answering competency question Q3: Concentration/Mass (enfosulfan sulfate)/Mass (wet flesh). It has been characterized as an Indirectly Measurable Element, as Concentration depends on sub-elements in order to be measured, addressing competency question Q5. According to NVS, concentration concept definition is “the amount of a specified substance in a unit amount of another substance or matrix”. The sub-elements to measure Concentration have been instantiated as Directly Measurable Element (elements that do not depend on others to be measured), answering competency question Q4.1: Mass (enfosulfan sulfate)/Mass (wet flesh).
It is important to highlight that the relation “characterized by” between Measurable Entity and Measurable Element is specialized from the homonym relation between Universal and Moment Universal, according to UFO. Thereby the directly measurable element (Mass) and the indirectly measurable element (Concentration) characterize the Measurable Entity (wet flesh) as the directly measurable element (Mass) characterizes the measurable entity (endosulfan sulfate), answering competency question Q6.
In other words, after applying the pattern TMElem, we could also identify different measurable levels for the compound concept variable used as an example, due to the relation between the patterns Measurable Entity and Measurable Element. First, we have identified wet flesh as the only measurable entity and then we instantiated it also to consider endosulfan sulfate as another measurable entity, representing a whole-part relationship between its parts. With this in mind, the competency question Q1 answer was updated to: wet flesh/endosulfan sulfate, as measurable entities for the compound concept variable. Likewise, Q2 answer was also updated to: Component of oyster anatomy/Pesticide.
In the Mea pattern, Measures are used for quantifying Measurable Elements and to characterize Measurable Entity Types. Measure is a Function in the sense that it maps an instance of Measurable Element to a value. TMea pattern characterizes a Measure into two types: Base Measure, which is functionally independent of other measures and used to quantify Directly Measurable Elements, and Derived Measure, which is defined as a function of other measures and used to quantify Indirectly Measurable Elements. Hence, Mass in Kilogram and Mass in Microgram were instantiated as Base Measure, answering competency question Q7, as they are used to quantify the directly measurable elements (flesh Mass and endosulfan sulfate Mass). Likewise, Concentration in Microgram/Kilogram is an instance of Derived Measure that quantifies the indirectly measurable element (Concentration), answering competency question Q5. Besides, the Derived Measure Concentration in Microgram/Kilogram is derived from the Base Measures Mass in Kilogram and Mass in Microgram, answering Q8.
In the Measurement Units & Scales group, important for the domain in order to model the variable units, the pattern MUnit was applied. As aforementioned, a quantitative variable might be expressed in different ways which requires units to be modeled independently of the associated variables. As units are essential information for describing measures, these concepts contribute for the interpretation of actual observations. In NVS, the concept “Concentration of endosulfan sulfate per unit wet weight flesh of Ostrea edulis” has as related concept “Micrograms per kilogram” that indicates the unit of weight of flesh wet mass and endosulfan sulfate mass, addressing Q9.
According to the application of these patterns’ groups, we emphasize how I-ADOPT ontology concepts and relations are semantically overloaded due to variable entities assuming different roles by the means of the relations they were associated with, e.g., ObjectOfInterest, ContextObject, or Matrix, and also due to the overload of the variable measurable entity and measurable element. Through the use of M-OPL, guided by the pattern’s competency questions, it is possible to clarify the core measurement conceptualization hidden for a compound concept variable, as identifying measurement entities and their properties that can be measured, besides defining measures and classifying them according to their dependence on others measures. Beyond being used to identify these core measurement concepts, they can be reused in several domains as they share some concepts in common, reinforcing that M-OPL patterns can contribute to this. 3.2.
According to I-ADOPT, an observable variable, as a concept that provides metadata for values made available in datasets, is a compound of at least one entity representing the ObjectOfInterest and one Property. These elements are represented in Figure 4 as Observable Quantitative Variable, ObjectOfInterest Entity and Property, respectively. Figure 4 highlights the relation between I-ADOPT ontology concepts, M-OPL patterns and UFO fragments.
The Property in a quantitative variable corresponds to the observed aspect, so it specializes the Measurable Element. The ObjectOfInterest Entity being the observable element to which the Property refers, specializes the Measurable Entity. The relate to relation is defined to represent the association of the Property with the ObjectOfInterest Entity. In addition to these elements, the Variable can optionally contain a Matrix, one or more ContextObject and one or more Constraint. These elements are represented in Figure 4 as Matrix Entity, ContextObject Entity and Constraint, respectively.
The Matrix represents the entity in which the ObjectOfInterest is contained. This information allows establishing a part-whole relation between Matrix Entity and ObjectOfInterest Entity whenever the variable presents a Matrix. Considering the part-whole relation, it also specializes Measurable Entity since it presents a measurable part. ContextObject refers to entities that provide context to the observation. Thus, in our model, ContextObject Entity is treated as a specialization of First OrderUniversal.
The relations established by the I-ADOPT ontology are also represented in Figure 4. Thus, the constrains relation associates Constraint with Matrix Entity, ObjectOfInterest Entity and ContextObject Entity. These restrictive aspects (Constraints) can define a specialization on the entity they constrain: the specialization of Matrix – flesh in wet flesh specifies exactly the flesh type considered by the observation.
The relations hasMatrix, hasObjectOfInterest, hasContextObject, hasConstraint and hasProperty are
also considered. These relationships associate the Observable Quantitative Variable with its constituent
elements. Intuitively, Observable Quantitative Variable emerges from its association with the other
concepts in the ontology. In UFO, it could be considered a Moment Universal Relator. It represents the
mereological sums of two or more externally dependent modes, i.e., of aspects of other individuals [
The relation between the measurable entity and the measurable element (Property) needs the context of the observation. In the instantiation shown in Section 3.1, it is observed that the Property (concentration) is a measurable element of the Matrix (wet flesh), derived from the measurable elements of the Matrix Entity and the ObjectOfInterest Entity.
By aligning I-ADOPT concepts with M-OPL and UFO fragments, it is possible to: (i) treat the semantic overload in the Entity concept, making explicit the interactions between them, e.g., the Matrix Entity contains the Object of Interest Entity; (ii) add new concepts referring to measurement, such as Indirect and Directly Measurable Element; (iii) provide new properties associating the established concepts to Functions, such as Measure used to quantify Property, and to Quality Regions, defining the Measure Unit in which the Variable Measure should be expressed. It is worth mentioning that these new elements are part of an ontological pattern based on UFO ontology, specifically acting with measurement aspects, regardless of the domain.
In this section, we present how M-OPL can be used to extend the I-ADOPT ontology concepts and relations, also capturing the conceptualization of measurement units & scales, measurement procedures, measurement planning, and measurement & analysis. In doing so, the I-ADOPT ontology can take advantage of these conceptualizations, semantically enriching the ontology schema level, to represent other important core measurement aspects from the compound concept variable, addressing complementary competency questions.
Figure 5 shows a fragment of the application of M-OPL pattern groups to extend the I-ADOPT ontology. It contemplates, besides the application of five M-OPL pattern groups presented in gray, other pattern groups highlighted in orange.
Considering the Measurement Units & Scales pattern group, the pattern MUnit&Scale is applied as it is also important to consider the variable scales of measures in the domain. According to MUnit&Scale pattern, measures can be expressed in Measure Units in which a measure is expressed as partitions of its scale. Measures have Scales composed of all possible values (Scale Value) to be associated by the measure to a measurable element. Reusing this pattern, the CQs 10, 11 and 12 could be addressed.
In the Measurement Procedures group, a Measured Procedure is applicable to a Measure. The MProc pattern is applied to model the steps to be carried out aiming a data collection for measures, addressing the CQs13 and 14. In the Measurement Planning group, the first applied pattern was INeed as it corresponds to information needs identified from measurement goals, addressing the CQ 15.
Measurement Goals are targets that can be used to guide the identification of the measures needed in a certain context. The next applied pattern was MPI-MP (Measurement Planning Item – Measurement Procedure) as it connects a Measurement Goal, an Information Need, a Measure, and a Measurement Procedure, meaning that the Measure meets the Information Need that is identified from the Measurement Goal.
Finally, the Measurement & Analysis group was applied. In the Meas (Measurement) pattern, it is possible to model data collection and analysis. Measurement is performed based on a Measurement Planning Item and it measures a Measurable Element of a Measurable Entity by applying a Measure and adopting a Measurement Procedure. The result is a Measured Value, which refers to a value of a measure scale. These could address CQs 16 and 17.
In summary, M-OPL contributes to address the conceptualization associated with observational variable measurement as it is organized in extensible modules according to specific measurement contexts. It can guide a common way of decomposing compound concept variables, identifying and reusing essential components as measurement entities, measures, measurement units & scales, measurement procedures, measurement planning and measurement & analysis.
OPLs facilitate the reuse of integrated ontological patterns into a conceptual model, leading to gains in reuse and the resulting ontologies quality. M-OPL addresses the core conceptualization for measurements in general and their characterization, organized according to an OPL. Measurement is very common in several domains as they share some concepts in common. An observational variable is a complex semantic representation of any type of data acquisition that usually carries ambiguous descriptions in human readable form or even for a machine. Especially for machine accessibility, it may be necessary to identify a concept with a standard label, providing a snapshot of the information associated with a particular compound concept variable.
The I-ADOPT framework ontology conceives a variable as a compound concept consisting of at least one entity (ObjectOfInterest) and one Property. Beyond that, other entities can be included to help contextualize the target object of observation. Although the ontology does not yet cover any additional concept or relation associated with units, instruments, measurable methods, time-related and geographical location information, these concepts provide essential information for interpreting actual observations. In particular, Units are essential information for describing measures and a quantitative variable might be expressed differently, requiring units to be modeled independently of variables.
This paper presented the use of M-OPL as a standard proposal for capturing the conceptualization of measurements for compound concepts variables in the I-ADOPT ontology. It also highlighted more specific aspects concerning measurement in high maturity representation levels. M-OPL addresses main measurements conceptualization that can be applied in several domains, as they share some concepts in common, contributing to create and to reuse complex unambiguous variable descriptions in a machine and human readable form. The benefits of using M-OPL to the I-ADOPT ontology are: (i) identification of concepts and relations semantically overloaded, since M-OPL patterns have been developed following a largely explored theory based on UFO; (ii) alignment to the core modeling patterns proposed for measurement, common to several application domains; (iii) capture of the conceptualization of measurements for compound concepts variables; and (iv) an extension to represent concepts not yet addressed by the ontology as scales and units, measurement procedures, measurement planning and measurement analysis. Besides the highlighted benefits, the I-ADOPT ontology could take advantage of the semantic richness of M-OPL to guide a common way of decomposing compound concept variables, identifying and reusing essential components as measurement entities, measures, measurement units & scales, measurement procedures, measurement planning and measurement & analysis.
As future work, we intend to use ontological patterns to represent qualitative variables and also explore new uses of M-OPL aiming to increase semantic interoperability of a larger set of complex concept variables. We believe this will contribute to avoiding inconsistencies and ambiguities between different variable interpretations, identifying main core concepts that are independent of the application domain.
This work has been partially supported with student grants from CAPES (Process numbers 223038.014313/2020-19 and 88887.613048/2021-00). 7. References Science and Computing, Stanford, ASE BIGDATA / SOCIALCOM / CYBERSECURITY Conference, 2014. [25] P. H. P. Richetti, and F. A. Baião, A Measurement Ontology for Beliefs, Desires, Intentions and
Feelings within Knowledge-intensive Processes, ONTOBRAS, 2018. [26] C. Fonseca, et al, Relations in ontology-driven conceptual modeling, in: International Conference on Conceptual Modeling, Springer, Cham, p. 28-42, 2019.