• No results found

Information Science and knowledge management practices, involves the develop- ment of tools and techniques for acquisition, representation, usage, preservation, as well as evolution of human knowledge. In order to use existing information to create knowledge, knowledge engineers need to understand and generate semantic relationships that are bound to exist between various terms, keywords and facets of domain knowledge. This can be made easier through the use of ontologies, which provides an explicit specification of terms, keywords or concepts in a given domain. These semantics, added to keywords using an ontology language, are human readable as well as machine process-able and thus gives an edge to using ontologies as tools for knowledge management. Moreover, in the article presented

in [33], which explore the roles of biomedical ontologies in knowledge manage- ment and data integration, the author further highlights the roles of ontologies in knowledge management systems (KMS) to include: annotation or indexing of resources, retrieval of data and information, data exchange and integration, pro- viding semantic interoperability among domain concepts, as well as knowledge discovery.

While biomedical ontologies are specifically mentioned in his article for do- main referencing, these roles are without doubt applicable to other domain ontolo- gies, including our underutilized crops domain. For example, we developed the UC-ONTO as domain ontology for underutilized crops that provide semantic re- lationships between crops-related terms and further integrate them with other crop data sources in RDFS to enhance information retrieval through a semantic search engine called the Onto-CropBase. Detailed description of the Onto-CropBase tool, including the comparable study semantic vs keyword-based searching, is presented in [11] and discussed here in Section 4.3.

2.3.1

Description Logics (DL) Based Knowledge Representa-

tion

The logical formalism behind ontological knowledge representation is known as description logic (DL). Being a decidable fragment of first-order predicate logic (FOL), DL is a collection of logic based knowledge representation formalisms designed for precise description and reasoning about the concepts in an applica- tion domain and the relations between them. DL uses logical symbols (operators, quantifiers, equalities, etc.) and variables combined with signatures of non-logical symbols such as unary predicates — which defines the domain concepts as Classes and the binary predicates — which define the number of individuals in a given class or the Roles and their multiplicity — which describes the relations between concepts. Typically, a DL axiom consists of these atomic concepts and individ- uals to represent complex knowledge of an application domain. As an example, consider a simple DL ontology described using axioms (2.1)-(2.3) that intuitively expresses the domain facts: ”Bambaragroundnut is a Legume Crop which has Leaf, Stem and Root as part of its Features and Leaf-spot is a disease of Bambara-

Groundnut”.

BambaraGroundnut v Crop u ∃isPartOf ·(Legumes u (∃hasFeatures · Features u

∀f eatures · (Leaf t Stem t Root))) (2.1)

LeafSpot ≡ Disease u ∃affects · Leaf (2.2)

BambaraGroundnut(BambaraGroundnutInd ) (2.3)

The ability to efficiently model a knowledge domain and the decidable com- putational characteristics, various ontology languages have been influenced by the DL syntax and semantics leading to the evolution of traditional ontology lan- guages, such as the Ontolingua, OKBC and F-Logic, among others [34]. With the introduction of the Semantic Web however, it became imperative to have a web ontology language with higher expressiveness. As mentioned in Section 2.2.2, this effort that leads to the serial development of various ontology markup lan- guages such as SHOE, RDF, RDFS, DAML and OIL, OWL 1 and more recently the OWL 2 language profiles.

DL-based knowledge representation systems usually involves two important components called the T-Box and A-Box. The ’T-Box’ or terminology-box con- tains the ontology concepts (owl:Classes) and Roles (owl:Properties) also called the terminologies. While the ’A-Box’ or assertion-box contains assertions of indi- vidual instances from the ontology terms. Example axioms in the T-Box could be the DL axioms (2.1) and (2.2) defined above in the simple Crop ontology, while a member of A-Box could be the third axiom (2.3) which asserts the individual ”BambaraGroundnutInd” into the ontology as a member of the ”BambaraGround- nut” class.

2.3.2

Classification of Ontologies

Ranging from generic taxonomies to specific application-level knowledge mod- els, Ontologies can basically be categorized into three [35, 36] namely: (i) The foundational ontologies, (ii) Domain ontologies and (iii) Application or Task-level ontologies.

Foundational Ontologies also called ’Upper-level’ or ’Reference’ ontologies, explain generic concepts and provide general taxonomies with multi-domain knowl- edge. The Unified Foundational Ontology (UFO) [37], Basic Formal Ontology (BFO) [38], General Formal Ontology (GFO) [39], and the GFO-Bio [40] among others, are common examples of foundational ontologies. Foundational ontology being a repository of general knowledge provides a means for semantic evaluation of lower ontologies such as the domain ontologies.

Domain ontologies on their part provide conceptual and more descriptive def- inition of terms within scoped domain boundaries, usually for an organization or knowledge community. They usually comprise of domain concepts, their rela- tionships and individual instances. They offer a common vocabulary for sharing, reuse and standardizing knowledge of a specific community or domain of dis- course. Larger domain ontologies are sometimes referred as upper-domain, such as BIOTOP [41], which is an upper-domain ontology for molecular biology link- ing smaller domain ontologies with the BFO, FAOs AGROVOC [42, 43, 44], which has in the past thirty years grown from simple multilingual agricultural index to a Linked-Open-Data (LOD) set. Other example domain ontologies in- cludes the Crop Ontology [4], Plant ontology [3], Gene Ontology [2]. For the purpose of our onto-cropbase tool [11], a domain-level ontology UC-ONTO [45] describing the underutilized crops and their farming practices, was developed to serve as knowledge base.

Application ontologies describe activities or specific application’s vocabulary in a domain. Developed to be used for specific applications, application ontologies usually utilize the domain ontologies by restricting its conceptualizations to model a specified application or task. For example, the Food Ontologies for nutritional applications in [46, 47, 48] and ’sensor ontologies’ for manufacturing application reviewed in [49]. The availability and popularity of standardized domain-level ontologies greatly affect the development of application-level ontologies, reduc- ing the process in most cases, to a simple task of narrowing down existing domain ontologies into task-specific ontologies. However, new domain/user-specific con- cepts can be easily generated from competency questions and user queries, when

involved in the knowledge generation process. These newly-generated concepts may need to be harmonized with the existing ones, leading to the Ontology stan- dardization process — see Section 2.5.4.

In the subsequent subsections, we briefly discuss some of the common ap- proaches of generating domain ontologies. These approaches, depending on the nature of the ontology development and knowledge engineers involved may be combined together or individually employed at different stages of the ontology development.