An e cient process for creating precise, accurate, machine-interpretable morphological descriptions of groups of organisms is needed to more e ectively gather observations of the world's biodiversity. While morphological descriptions are required for the publication of modern scienti c names, it is common for these descriptions to get revised after the initial publication which can lead to data loss. A system for creating and naming machine-interpretable descriptions of groups of organisms, the Semantic Vernacular System, is proposed as a solution for creating such descriptions and managing their relationship to formal scienti c names while improving the collection of observational data.
as the `type specimen'. For all names, there is an assumption that all the organisms included are
part of a single evolutionary lineage. A formal description or `circumscription' is required for a
name to be validly published. The circumscription is intended to separate the described group
of organisms from those described by other names at the same level. The primacy of the type
specimen and the evolutionary lineage is demonstrated by the frequent revision of circumscriptions
when new evolutionary evidence is found with respect to the type specimen. For example, the
circumscription of Laetiporus sulphureus, the Sulphur Shelf Mushroom, was signi cantly reduced
when mating studies showed that the original circumscription included at least 5 distinct species
[
While many biodiversity observation systems encourage photographs and notes to document an observation, frequently a scienti c name is simply asserted with no explicit evidence. When just a name is provided, there is an immediate loss of information for any name that has multiple circumscriptions in current use. For example, if an occurrence of Laetiporus sulphureus is recorded, it is not clear which existing circumscription was intended. This loss would be further compounded by any further revisions unless the users of that data are careful to keep track of when new circumscriptions are created relative to each observation.
One possible approach would be for biodiversity observation systems to require that users provide
an explicit reference to the circumscription they used when making an identi cation. Sites such as
the Encyclopedia of Life (EOL) [
A better approach from a data management standpoint would be to require a person making an identi cation to be explicit about the features they based their identi cation on. However, given the lack of even standardized terminology for many groups, this in e ect means the identi er would have to write a detailed description for each identi cation. Validation would be better supported with this approach, but would be very subject to the consistency and thoroughness of the identi er. Both of these approaches would frequently impose a signi cant overhead burden on the taxonomic experts. 3
The proposed alternative is to create the Semantic Vernacular System which enables authoring
named, machine-interpretable de nitions of groups of organisms that are then associated with sets
of scienti c names. The Web Ontology Language [
Just as with scienti c names, it would be valuable for the new system to be peer-reviewed, strictly prevent the reuse of names, apply any agreed upon nomenclatural rules, and avoid the unintended re-publishing of an existing description. The Semantic Vernacular Descriptions, or SVDs, envisioned will be `born digital' in a freely available, global repository. This will allow all of these desirable features to be applied consistently and automatically from the ground up. Once an SVD has been approved through the peer-review process the association of the chosen name with that semantic description will be considered a strict de nition that should never change just as the association between a scienti c name and its type specimen should never change. This tight association means that a particular SVD applies to any observed organism that matches that semantic description. SVDs will allow users to apply precise sets of features by name to their observations. Biodiversity observation systems using the system will allow users to automatically review and even explicitly con rm the de ning features for any name they apply. While there will, of course, be human error in the application of such names and even speci c features, the data recorded by the observer will be explicit and will not degrade over time. For Laetiporus sulphureus this means separate names for the historical circumscription as well as names for each morphologically distinct group whether it includes multiple species, a single species, or a distinctive subset of one or more species.
The proposed system will also support the registration and naming of descriptions that
correspond to common observational experiences. This will allow users to record what are in e ect
partial identi cations to groups of `look alikes' that may or may not include all the members of
an evolutionary lineage. Currently groups of species that are di cult to tell apart in the eld,
sometimes referred to as `sibling speices' [
Populating such a system with meaningful descriptions will not be simple. It must start with the
development of ontologies that capture and standardize the terminology needed to describe observed
features. This work has begun in some groups such as Teleost sh [
A demonstration system [
Finally, the proposed machine-interpretable de nitions naturally lead to a novel system for computer-aided identi cation of observations. As users learn to describe their observations in the same way that the descriptions are stored, it will be straight-forward for the system to indicate what existing SVDs match the given features and the implied set of potential scienti c names. This position paper describes the need for a new class of names tightly associated with semantic descriptions of groups of organisms. We outline the creation of the Semantic Vernacular System for managing these new names and descriptions. This system will enable more precise and accurate observations of biodiversity with minimal additional overhead while encouraging the creation of machine-interpretable descriptions with clear connections to traditional scienti c names.