Knowledge Representation is important part of AI. The purpose is to reveal best possible representation of the Universe of Discourse (UoD) by capturing entities, concepts and relations among them. With increased understanding of various scientific and technological disciplines, it is possible to derive rules that governs the behaviour and outcome of the entities in the UoD. In certain cases, it is not possible to establish any explicit rule, yet through experience or observation, some experts can define rules from their tacit knowledge in specific domain. Knowledge representation techniques are focused on techniques that allows externalization of implicit and explicit knowledge of expert(s) with a goal of reuse in absence of physical presence of such expertise. To ease this task, two parallel dimensions have developed over period of time. One dimension is focused on investigating more efficient methods that best suit the knowledge representation requirement resulting in theories and tools that allows capturing the domain knowledge (Brachman & Levesque, 2004). Another development has taken place in harmonization of tools and techniques that allows standard based representation of knowledge (Davies, Studer, & Warren, 2006). Various languages are proposed for representation of the knowledge. Reasoning and classification algorithms are also realized. As an outcome of standardization process, standards like DAML-OIL (Horrocks & Patel- Schneider, 2001), RDF (Manola & Miller, 2004) and OWL(Antoniou & Harmelen, 2004) are introduced. Capturing the benefit of both developments, the tooling is also came in to existence that allows creation of knowledgebase. As a result of these developments, the amount of publicly shared knowledge is continuously increasing. At the time of this writing, a search engine like Swoogle (Ding et al., 2004)-developed to index publicly available Ontologies, is handling over 2,173,724 semantic web documents containing 431,467,096 triples. While the developments are yielding positive results by such a huge amount of knowledge available for reuse, it have become difficult to select and reuse required knowledge from this vast pool. The concepts and their relations that are important to the given problem could have already been defined in multiple Ontologies with different perspectives with specific level of details. It is very likely that to get complete representation of the knowledge, multiple Ontologies must be utilized. This requirement has introduced a new discipline within the domain of knowledge representation that is focused on investigation of techniques and tools that allows integration of multiple shared Ontologies.
The problem of Ontology integration is not completely new. Schema Matching is a similar problem being addressed in the context of enterprise integration. But, in Ontology matching, the scale and complexity is much higher and requires special considerations. (Shvaiko & Euzenat, 2006) highlights the key similarities and differences between both the techniques. In schema matching, the semantics of the given term is guessed whereas the ontology matching methods relies on deriving the semantics from explicit representation of concepts and relations in given Ontology. Numerous methods and approaches have been proposed that attempt to solve the problem targeting specific aspects of the represented knowledge(Ehring, 2007).
Apart from standards that guide the languages used for the development of Ontology, some standard Ontologies have also been defined. The role of these Ontologies is to provide framework of vary basic elements and their relations, based on which complex domain knowledge can be developed. SUO(Niles & Pease, 2001), SUMA(Niles & Pease, 2003), OpenCyc(Sicilia et al., 2004) are examples of the same. SWEET (Raskin, 2003) provides standard Ontologies in environmental science domain. Hence, the levels in Ontology also address important dimension in knowledge engineering through integrating available Ontologies.
Key Terms in this Chapter
Ontology Alignment: Ontology Alignment is a process to articulate similarity in the form of one-to-one equality relation between every elements of two separate Ontologies.
Ontology Merging: Ontology Mapping is a process that results in generation of a new ontology derived as a union of two or more source Ontologies of same subject domain.
Ontology Integration: Ontology Integration is a process that results in generation of a new ontology derived as a union of two or more source Ontologies of different but related subject domain.
Context Aware Techniques: Techniques that are focused on nearness of weight assigned to specific relations among concepts considering application context as basis for mapping.
Structure Aware Techniques: Techniques that also consider structural hierarchy of concepts as basis of mapping.
Extension Aware Techniques: Techniques that are focused on finding nearness among features of available instances of different Ontologies to form basis for mapping.
Ontology Mapping: Ontology Mapping is a process to articulate similarities among the concepts belonging to separate source Ontologies.
Intension Aware Techniques: Based on the Information flow theory, techniques that are focused on finding two different tokens (instances) belonging to separate Ontologies that maps to single type (concept) as a basis for mapping.
Articulation Ontology: Articulation ontology consists of concepts and relations that are identified as link among the concepts defined in two separate Ontologies also known as articulation rules.
String Similarity Techniques: Set of techniques that uses syntactic similarity of concepts as basis of mapping.
Semantic Similarity Techniques: Techniques that are focused on logic satisfiability as basis of mapping.
Linguistic Similarity Techniques: Set of techniques that refer linguistic nearness of concepts in the form of synonyms, hypernyms, and hyponyms by referring to related entries in the thesaurus as basis for mapping.
Ontology Mediation: Ontology mediation is a process that reconciles difference between separate Ontologies to achieve semantic interoperability by performing alignment, mapping, merging and other required operations.