More GBIF questions (was: ITIS)

[Charles Hussey]

This is proving a very stimulating thread that demonstrates the value of
TAXACOM - would that more of the people posting could make it to TDWG
meetings (www.tdwg.org) as it would be good to be able to talk face to face.

I thought that it might be worth mentioning some of the experience that we
are gaining from managing the National Biodiversity Network Species
Dictionary as this is, in a way, a microcosm of GBIF's ECAT (rather than
Catalogue of Life/ITIS). The Dictionary aspires to become the master list of
what is found in Britain and exists to deliver authoritative lists to the UK
recording community, supports taxonomies used by Government Agencies (such
as the Environment Agency), and informs decision-makers. It is also there to
support other NBN initiatives such as the NBN Gateway, which is now serving
access to over 15 million observation records (www.searchnbn.net) - and
which will shortly be searchable through the GBIF portal.

The Dictionary follows the NBN "classic" data model, developed by Charles
Copp
(http://www.bgbm.org/biodivinf/docs/archive/Copp_C_2000_-_NBN_Data_Model.pdf
), which Charles has now evolved and transformed to become the BioCASE
thesaurus model
http://www.biocase.org/Doc/Results/WP4/D9CompleteDraftThesaurusModel.pdf.
The Dictionary is able to handle taxon concepts (using TAXON_VERSION and
TAXON_VERSION_RELATION tables).

The Dictionary is assembled from a collection of checklists (over 170 to
date). Formerly these were published (static) lists but we are now starting
to incorporate dynamic (maintained) checklists. Many of the checklists,
although of value in terms of coverage, may lack an internal hierarchy and
may omit authorities for names. Given that we have many overlapping lists,
we have assembled a great confusion of names (as, I expect, will any project
that attempts to search across multiple datasets) and, for groups that we
have examined in detail, we are finding the ratio of well-formed recommended
names to all other names (vernaculars, synonyms, incomplete names and
orthographic variants) is 1:9 (i.e. only 10% are valid). We don't edit
names! - we aim to be able to deliver a given checklist verbatim (besides
which, ownership of the data resides with the list providers). So, we are
finding that the provision of a name-server service is essential to provide
coherence to our data sets. We can only do this for groups where we are
already in possession of a comprehensive, authoritative and maintained
checklist. We can then map all names in the Dictionary belonging to that
group (finding them all is a challenge in itself) to a recommended name.
This makes it possible to enter any name, find its recommended name and then
work backwards to find all other associated names; thus providing a vital
query-expansion tool.

Some specific points that I should like to make are:
1) Bringing names from many sources together into a "data warehouse" greatly
facilitates error detection (which can be fed back to the data providers.
2) Mapping names to build a name-server is resource intensive and cannot be
fully automated. Many mappings are relatively obvious, but some will always
require expert scrutiny (Many people are involved in helping to build the
Species Dictionary. We shall be encouraging online feedback to report errors
and omissions, but these will be referred to owners of the maintained lists
and all editing is currently done off-line).
3) Whilst publicly accessible names services (such as the GSDs with full
synonymy provided by Species 2000) are very valuable resources, we have
found that when linking to "real" data sources (unit level data sets of
specimen or observation records), the name-server has to incorporate all
manner of names that transgress rules of nomenclature yet which have to be
included if accurate results are to be returned in searches.
4) Hence, in the Dictionary, "erroneous" names are not discarded - they may
actually be in common usage amongst recorders (some bat names spring to
mind). Importantly, they are often out on the web and need to be included in
search strings.
5) It is important (to my mind) that the developing schemas allow assertions
to be made about names (so-and-so said, on such-and-such a date, that this
version of a name is well-formed and represents current usage in Britain).
We are investigating aspects of the semantic web (RDF and ontologies) to
support this.

I hope to be able to present more on our name-server work at the CODATA
Conference in November (http://www.codata.org/04conf/).

PS I hesitate to recommend the Dictionary website to you
(www.nhm.ac.uk/nbn/) - it is in need of an overhaul and does not adequately
represent the current state of the Dictionary project.

Charles Hussey,

Science Data Co-ordinator,
Data and Digital Systems Team,
Library and Information Services,
The Natural History Museum,
Cromwell Road,
London SW7 5BD
United Kingdom

Tel. +44 (0)207 942 5213
Fax. +44 (0)207 942 5559
e-mail c.hussey at nhm.ac.uk
Species Dictionary project: www.nhm.ac.uk/nbn/
Nature Navigator: www.nhm.ac.uk/naturenavigator/


Charles Hussey,

Science Data Co-ordinator
Library and Information Services

Tel. Ext. 5213
Fax. Ext. 5559
e-mail c.hussey at nhm.ac.uk