An Ontology Blueprint for Constructing Qualitative and Quantitative Scienti c Variables

Abstract

This work presents an ongoing e ort to develop simple ontological design patterns for describing scienti c variables with a high level of speci city in resource description format (RDF). The application of the ontology design patterns discussed here were used to create a variables ontology for the geosciences. The long-term aim of this work is to develop an ontological blueprint for automated ontology generation from a corpus. Such ontologies can be used for semantic mediation in automated scienti c work ows and semantic alignment of content in heterogeneous resources.

Keywords

• ontology design pattern scienti c variables semantic mediation

Introduction

The Ontology for Constructing Scienti c Variables (OSV) is a mechanism for storing conceptual information necessary for identifying, disambiguating, and assembling scienti c variables. OSV is a successor to the Geoscience Ontology (GSN)[ 4 ] and extends the principles introduced in the CF standard names [ 2 ] and the CSDMS standard names (CSN) [ 3 ]; whereas the aforementioned naming e orts relied on encoding scienti c variables using controlled vocabularies and one-dimensional strings, the OSV is terminology-agnostic and encodes relational and contextual information via the Resource Description Format (RDF), resulting in a richer representation with more degrees of freedom. OSV is a critical tool for semantic mediation, providing the language to link unstructured information contained in large corpora to structured information captured in data sets and used by computational models. Along with other interpretative tools, OSV is designed to enable automated alignment and integration of distributed scienti c information.

There are a wide range of scienti c ontologies available, see e.g., [ 5,6,7 ]. However, although these ontologies are useful for speci c applications, there is, to the authors' knowledge, no available ontology that (a) provides the desired specicity for distinguishing variables at a highly granular level within a domain, (b) comprises patterns that are readily extensible to other domains, and (c) denes mandatory components of a variable. The ontology we present in this work aims to decompose and modularize the construction of scienti c variables, explicitly labeling required elements that must be provided in order to completely and unambiguously identify the concepts represented by a scienti c variable| namely an object of observation, a corresponding property, and a quantity with units. We start by identifying the core ontology building blocks in Section 2 and then describing how the building blocks are combined to build complex concept representations. 2 2.1

Concept Class De nitions

Physical Concepts A Phenomenon is a fact or situation that is observed or could be observed to exist or happen in the physical world. A phenomenon that is observed to exist is at equilibrium, whether dynamic, chemical or static, and one that is observed to happen is removed from equilibrium, experiencing a change of state as a result of certain processes. A phenomenon consists of the substance of which it is made (Matter), a Form that de nes its occupation of space, and possibly, one or more Processes. Phenomena are de ned recursively[ 1 ], where any given phenomenon can be decomposed into smaller phenomena and can be combined with other phenomena to build larger, more complex phenomena. A Body is a phenomenon at equilibrium that is identi ed by its Matter and Form. A Process is a set of actions that may occur in parallel or sequentially. 2.2 A System is the abstracted, diagrammatic representation of a phenomenon, and includes any applied, human-contrived physical or mathematical abstractions or models. In OSV, a system that has a relatively unchanging state is static, while a changing system is dynamic. A static system may comprise multiple dynamic systems which together are at equilibrium. Like Phenomena, Systems are de ned recursively.

A Property is a characteristic or feature of a system. A Value may be numerical or categorical and represents a system state, evaluated either objectively or subjectively; it is associated with a property. A Quantity is a numerical value with associated units. An Attribute is a property-value pair. It is important to note that some properties may be observable but may not be able to be measured directly and may be assessed through manipulation of other attributes; examples include severity and resilience.

A Variable is a phenomenon-property pair. It must comprise an object of measurement|one or more Phenomena|as well as a Property. As an example, `precipitation' is not a complete variable, as it only identi es a process, and neither is `rainfall', as it only identi es a phenomenon|the precipitation of water from clouds. In order to properly identify a variable, a property (such as `volume ux' or `duration' in the case of rainfall) must also be identi ed.

Building a Variable

The steps for identifying the components of a scienti c variable are: 1. Select a phenomenon of interest for study{this is called the object of observation and will be the object of the variable. 2. Select one or more properties of that phenomenon to evaluate. 3. Diagram that phenomenon for the desired analysis, and if necessary, identify any applied abstractions, such as approximate mathematical or physical models (e.g., surface, ellipsoid, etc.).

A system is de ned recursively in the ontology and comprises one or more participants, the role of each participant, and accompanying processes. Participants are recursively de ned as distinct subsystems of the larger whole to provide the desired level of granularity. The granularity of any system may be further re ned by identifying system attributes (system state) that are constant for the scope of measurement.

Figure 1 provides an overview of the di erent systems that can be modeled. Static systems involve processes that are at equilibrium while dynamic systems are removed from equilibrium. The single-body, static system is equivalent to the Body class. Matter is a type of multiple-body, static system. When enclosed with a boundary, a multiple-body, static system may be turned into a static, single-body system. When a Form is applied to Matter, a Body system results.

Single Body

Multiple Body

c i t a t S c i m a n y

D

A variable is assembled by linking the system of interest to the desired property. If applicable, a variable may also include a reference frame for the evaluation of the property, as well as context phenomena. Figure 2 shows an example of how the building blocks are used to build a variable.

fuel attribute: gaseous participant role: consumed consumption

_:? participant role: consumer participant role: source emission carbondioxide participant role: main mass World Development Indicator: CO2 emissions from gaseous fuel consumption (kt)

GSN construction: carbon-dioxide~emitted-from-fuel~gaseous-consumption_mass

4 Implementation

The Geoscience Ontology[ 4 ] is an example of a domain-speci c OSV application which expresses a wide range of scienti c variables. The linked website provides a web interface to a SPARQL endpoint to query a beta version of the ontology.

References

1. Hooft , G.: In Search of the Ultimate Building Blocks . Cambridge University Press, Cambridge, UK ( 1997 )

2. Guidelines for Construction of CF Standard Names , http://cfconventions.org/ Data/cf-standard-names/docs/guidelines.html. Last accessed 1 June 2018

3. CSDMS Standard Names , https://csdms.colorado.edu/wiki/CSDMS_Standard_ Names. Last accessed 1 June 2018

4. Geoscience Ontology , http://www.geoscienceontology.org. Last accessed 1 June 2018

5. SWEET Ontology , https://sweet.jpl.nasa.gov/. Last accessed 24 July 2018

6. QUDT Ontology , http://www.qudt.org/. Last accessed 24 July 2018

7. GCOS Ontology , http://vocab-test.ceda.ac.uk/ontology/gcos/ gcos-content/. Last accessed 24 July 2018