SKOS Sources Transformations for Ontology Engineering: Agronomical Taxonomy Use Case

Abstract

Sources like thesauri or taxonomies are already used as input in ontology development process. Some of them are also published on the LOD using the SKOS format. Reusing this type of sources to build an ontology is not an easy task. The ontology developer has to face different syntax and different modelling goals. We propose in this paper a new methodology to transform several non-ontological sources into a single ontology. We take into account: the redundancy of the knowledge extracted from sources in order to discover the consensual knowledge and Ontology Design Patterns (ODPs) to guide the transformation process. We have evaluated our methodology by creating an ontology on wheat taxonomy from three sources: Agrovoc thesaurus, TaxRef taxonomy, NCBI taxonomy.

Dealing with Incompatibilities During a Knowledge Bases Fusion Process

“…The Discovery of the Optimal Extension allows the generation of an extension representing the maximal subset of compatible candidates validated by an expert (optimal extension). The first three activities have been already presented in [1,2]. In this article, we focus on the discovery of the optimal extension.…”

“…More information about this project are available in [1,8]. In this evaluation we only take care of the creation of a wheat taxonomy.…”

“…These sources have been transformed in knowledge bases using our SKOS transformation method [1], based on ontological module to drive the transformation process and define transformation patterns dedicated to each source. For our experiment, we use the ontological module called AgronomicTaxon 5 that merges several ontology design patterns [8].…”

“…Our previous works present several functions to compute the trust score [1]. These scores can be used to optimize the objective function of our extension generation algorithm.…”

“…For each candidate, we evaluate its trust score. Previous work [1] propose several trust measures, which evaluate consensus found in existing KBs. Merging different KBs about the same domain may generate errors since the KBs can propose different viewpoints on a same domain.…”