Biological Taxonomy and Ontology Development: Scope and Limitations

Abstract

Abstract.  The prospects of integrating full-blown biological taxonomies into an ontological reasoning framework are reviewed. Traditionally ontological representations of taxonomy have adopted the model of a single and static hierarchy. This model is contrasted with a more realistic situation involving dynamic revisions of particular groups and alignments among alternative taxonomic perspectives. Taxonomic practice is bound by a range of epistemological constraints and linguistic conventions that run orthogonal to the logical background from which ontological entities and relationships originate, resulting in severe challenges for ontological representation and reasoning. In particular, the purported existence of a single hierarchy in nature forces taxonomists to gradually approximate this hierarchy and make frequent rearrangements in light of new evidence. The evolvability of taxa implies that taxon-defining features may be lost in subordinate members or independently gained across multiple sections of the tree of life. As a result, many terms for phenotypic properties are phylogenetically underdetermined and have limited hierarchical transitivity. The standard approach of defining taxa both in reference to properties (intensional) and members (ostensive) undermines the individual/class dichotomy that sustains conventional ontologies, and compromises the reasoning capabilities associated with this distinction. Neither the use of Linnaean ranks in taxonomy nor the 250-year legacy of nomenclatural adjustments have obvious analogues in the ontological realm. Lastly, the piece-meal appearance of full-blown taxonomic information makes it necessary to use expert alignments to obtain a comprehensive static perspective. In light of these limitations, research along the taxonomy/ontology interface should focus either on strictly nomenclatural entities and relationships or on ontology-driven strategies for aligning multiple taxonomies, but not on building static networks for large portions of the tree of life. The prospects of using ontology-based services in taxonomy will largely depend on the ability of the taxonomic expert community to present its products in ways that are more compatible with ontological principles than concurrent practice.

Names are not good enough: Reasoning over taxonomic change in the Andropogon complex1

“…They are therefore subject to adjustments whenever new evidence regarding the identity of taxonomic entities or relationships among these is brought forth by the latest systematic research [37]. For many organismal groups in the tree of life, systematists are not close to completing this process of adjustment.…”

“…Typically both a type and a feature-based circumscription are provided when anchoring the meaning (referential extension) of a taxonomic name [29,34,37,96]. However, the former arbiter -i.e., the type identity -has special weight when dealing with alternative name:meaning (read: "name-to-meaning") assignments that become necessary when taxonomies undergo revisions.…”

“…This approach offers an immediate and valuable service to users. However, in the longer term it often leads to multiple distinct perspectives being represented by earlier and later versions of the 'same' standard [5,37,90]. Thus in effect the unitary taxonomy turns into an open-ended temporal chain of partially incongruent taxonomies.…”

“…The logic foundations for this particular approach were developed in [16,43,86]. The step of modeling an input taxonomy as an is_a hierarchy is well established [37,66,89]. However, the remaining steps in our toolkit workflow diverge from existing ontology matching or provenance-tracking applications [18,21,24,52,80,89,92,97].…”

Knowledge acquisition: Past, present and future

“…A nice example of the issues involved is to be found in research on merging what appear to be well-defined biological taxonomies of the same species from different sources (Thau and Ludäscher, 2007), or of merging them across species to create an overall tree of life (Bininda-Emonds, 2004). Pioneers in this area have become disenchanted by the lack of consistency in what initially appeared to be a logically and conceptually well-founded discipline and have documented in some detail the ways in which in practice the taxonomists have muddled through in a way that currently undermines attempts to create computational ontologies (Franz and Thau, 2010).…”