Ontology Engineering with Rough Concepts and Instances

Abstract

Abstract. A scenario in ontology development and its use is hypothesis testing, such as finding new subconcepts based on the data linked to the ontology. During such experimentation, knowledge tends to be vague and the associated data is often incomplete, which OWL ontologies normally do not consider explicitly. To fill this gap, we use OWL 2 and their application infrastructures together with rough sets. Although OWL 2 QL is insufficient to represent most of rough set's semantics, the mapping layer of its Ontology-Based Data Access framework that links concepts in the ontology to queries over the data source suffice to ascertain if a concept is rough, which subsequently can be modelled more precisely in an OWL 2 DL ontology. We summarise the trade-offs and validate it with the HGT ontology and its 17GB genomics database and with sepsis, which demonstrates it is an encouraging step toward comprehensive and usable rough ontologies.

The African wildlife ontology tutorial ontologies

“…Finally, the AWO can be used on its own or together with the textbook "An Introduction to Ontology Engineering" [16], which contains examples, tasks and exercises with the AWO.…”

“…In the case of the AWO, this is not specifically with respect to OWL language features, but one of notions of ontology quality and where one is in the learning process. For instance, version 1a contains answers to several competency questions-i.e., quality requirements that an ontology ought to meet [17]-that were formulated for Exercise 5.1 in the "Methods and methodologies" chapter of [16]. Versions 2 and 3, on the other hand, have the AWO aligned to the DOLCE and BFO foundational ontologies, respectively, whose differences and merits are discussed in Chapter 6 of the textbook, ensuring discussion of refinements in ontological precision with, e.g., processes and dispositions (e.g., an Eating class with participating objects cf.…”

Modeling Uncertainty in Ontologies using Rough Set

“…He experimented his approach with rough concepts and vague instances using the HGT ontology with the HGT-DB database and septic patients. [23] is the same as [24] except that the language considered is OWL2.…”

Rough subsumption reasoning with rOWL

“…Compared to fuzzy logics where one already knows the properties and fine-tunes their values, rough set applications enable one to experiment with finding the optimal set of properties of a set of objects in the software system. This approach has the potential to be very useful also in knowledge management, as demonstrated by promising implemented use cases for in silico hypothesis testing with bio-ontologies as part of a scientist's research methodology [10] and disease characterisation and patient classification using electronic health record data [19]. To realise rough knowledge management, rough sets have to be integrated with the knowledge representation layer and suitable reasoning services need to be devised with the aim to prove that the rough knowledge is not only consistent but that one also can avail Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.…”

“…There are several formalisms of rough sets, such as in Datalog ¬ [4], extended logic programs [21], and Description Logics (DL) with extensions to the Web Ontology Language OWL in particular [2,5,7,8,12,13,19], where each language proposed includes core notions of rough sets only in part [9] and it is not optimised for the most expressive and recently standardised OWL 2 ontology languages that already enjoy substantial implementation infrastructure and user uptake. Reasoning with rough ontologies falls into two categories: reasoning over the instances by querying the data to ascertain if the class is indeed a rough class [10] and type-level reasoning, with the principal reasoning services being possible and definite satisfiability, and rough subsumption reasoning and classification of the rough classes. Thus far, only [8] considers possibly and definitely satisfiable and crisp subsumption of rough concepts for an arbitrary DL language, RDL AC , but neither what can be deduced about the rough concepts from their respective approximations nor enforced on the approximations given an asserted subsumption of rough concepts.…”