Hi Peter et al.,
You should look at this paper *A Gross Anatomy Ontology for Hymenoptera* http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0015991
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0015991It is an example of how one might create an ontology that helps clear up the meaning behind all these anatomical terms etc.
One of the things that struck me was we seem to think along similar lines but in different domains.
*"Though counterintuitive to some, the development of anatomy ontologies proceeds more effectively without reference to homology. Circumscriptions of classes in an ontology must first and foremost (at least within the goals of the HAO, and we feel in general) allow for the identification of instances of the class in question. These circumscriptions are crudely analogous to engineering blueprints in that they allow a domain expert to identify, with reference to an individual, some instance of a concept (e.g. the anterior ocellus on the specimen identified with the identifier NCSU 1234). Another central reason for decoupling anatomy from phylogeny (homology) is that it maximizes the potential cross-domain application of the ontology."*
In other words, the concept comes first and then you attach names to that concept.
Respectfully,
- Pete
On Mon, Jan 10, 2011 at 5:17 PM, Peter Stevens peter.stevens@mobot.orgwrote:
To Donat's message also - absolutely, and it is because we cannot do this that we are floundering so much. Because all we have are assertions - that some dignify as "hypotheses" - that such and such specimens go in a species (and in many floras you don't get even that), taxonomists simply cannot readily build on the past. When the measurements we make are linked to the specimens we observe, then we will be in business. A solution is to be a neo-Linnaeus - i.e., we need to enable things like names in current use that allow us to ignore earlier literature and the headaches that interpreting it cause us, or we can go to that literature for any useful information it may contain, but it will not have any nomenclatural implications.
OK, I know I am glossing over a great deal. I am not suggesting that the only people who can be neo-linnaeans are those who have carried out multivariate analyses and have their data in a public repository - most of us, myself included, do not carry out such analyses, not are they necessary in all cases. Digitisation of all literature is great, but why i as a taxonomist should have to go back in perpetuity to check out whether Zollinger or Miiquel's concept of a species is the same as mine, let alone what they actually meant by lanecolate, is beyond me (unfortunately, Vernon Heywood has pre-empted the sisyphean title). Ditto about image digitization, but I wait for the automatic recording of simple morphological data from these.
I have to have all this worked out in six weeks or so for Berlin....
P.
On Jan 9, 2011, at 9:25 AM, Peter DeVries wrote:
Hi Donat,
On Sun, Jan 9, 2011 at 4:55 AM, Donat Agosti agosti@amnh.org wrote:
A simple solution would be that in future to each usage of a name in taxonomic literature, at least for treatments, the name has to be explicitly linked to the underlying specimen or observation data. Since this is part of a treatment, the identifier should allow to retrieve the treatment and through it the specimen data or at least the observation data and the author who used it in this specific selection.
This is the goal I have been trying to achieve with TaxonConcept.org.
An open, resolvable, identifier that provides some information as to what the entity is.
With the use of namespaces these can be short and easy to use.
Puma concolor se:v6n7p - the current scientific name followed by the asserted concept.
They are currently not as informative as I would like, but allow linked descriptions, revisions, type specimens, DNA etc.
Eventually the curation of these concepts would be done by taxon experts, who should receive some sort of academically meaningful credit.
This credit could be easily tracked via Linked Open Data identifiers.
As I have said many times on the list, I would like to work with out in conjunction with GBIF, EoL and the rest of the community.
I was not involved in the planning or development of the GNI, my role was simply to test ways of connecting various names to some sort of "concept" using the Linked Open Data approach. At the TDWG meeting we were able to show that this could be used to create a queryable synonymy knowledge base.
As I have said before, I think it is best to start with concepts and then connect that concept to the various related names.
Like Rod, I also like uBio and I am sure he would support moving those resources to an updated vocabulary that does not use LSID's.
Respectfully,
- Pete
Donat
Why aren't identifiers reused?
Because in most cases they offer no added value. If I have a ITIS TSN there's not much I can do with it. I can get a name from ITIS (with some vague assurance that thus name is accepted - or not - with no evidence for this assertion). I can think of only two taxonomic identifiers that have real value and get much reuse, NCBI taxonomy ids and uBio NameBankIDs. NCBI ids are reused because they underpin the genomics databases, and genomics does real computational biology, and makes extensive reuse of data (as exemplified by the annual Nucleic Acids Research database issue). uBio NameBankIDs get reused because uBio has lots of names, and provides services for discovering those names in text (see, e.g., their use by BHL).
Few taxonomic name databases provide a compelling reason for anyone to use their identifiers, most being digital sinks (you go there, get an identifier for a name, and nothing else).
Why UUIDS?
UUIDs are ugly, and solve a problem that for the most part we don't have. They are ideal for minting globally unique identifiers in a distributed system, but we don't have distributed systems. Catalogue of Life uses UUIDs, but these are centrally created (I suspect using MySQL's UUID function, given how similar the UUIDs are to each other). ZooBank uses them, but it is not (yet) a distributed system. If the Catalogue of Life were genuinely a distributed system UUIDs would make sense, but that's not actually how it works.
I think users would cope with UUIDs if the databases using them provides clear value. For example, MusicBrainz uses UUIDs http://musicbrainz.org/artist/563201cb-721c-4cfb-acca-c1ba69e3d1fb.html , as does Mendeley, the latter hiding them from users via human- readable URLs. Given that we have obvious user-friendly candidates for URLs (taxonomic names), it would be trivial to hide UUIDs in names (making homonyms distinct by adding authorship, or whatever it took to make them unique as strings).
What, if anything, is a taxonomic name?
In my experience, when non-taxonomists meet taxonomists things get ugly. For example, a publisher wanting to mark up taxonomic names in text might ask taxonomists how to do this,and within minutes the taxonomists are off into discussions of namestrings versus usages versus concepts and pretty soon the publisher deeply regrets ever asking the question. I've been at meetings where the look in publishers' eyes said "run away, run away".
Part of the reason we have multiple databases is because different projects are capturing different things (roughly speaking, uBio is mostly about namestrings, Catalogue of Life is about concepts, IPNI and ZooBank are about first usage of a name, etc.)
Most users outside our field won't give a damn about the niceties of these distinctions, yet we persist in discussing them ad nauseam. Until we provide a single, very simple service that takes a name string and hides all this complexity (unless the user chooses to see the gory details) while still providing useful information, we will be stuck in multiple identifier hell. The tragedy is we've never had more people genuinely interested in linking to names than at present -- publishers are desperately trying to add "semantic value" to their content, and we are spectacularly ill-equipped to deliver this (and it's our own fault).
I rather suspect we're rapidly approaching the point where users outside taxonomy will simply say "to hell with these taxonomists, let's just use Wikipedia and be done with it."
Regards
Rod
Roderic Page Professor of Taxonomy Institute of Biodiversity, Animal Health and Comparative Medicine College of Medical, Veterinary and Life Sciences Graham Kerr Building University of Glasgow Glasgow G12 8QQ, UK
Email: r.page@bio.gla.ac.uk Tel: +44 141 330 4778 Fax: +44 141 330 2792 AIM: rodpage1962@aim.com Facebook: http://www.facebook.com/profile.php?id=1112517192 Twitter: http://twitter.com/rdmpage Blog: http://iphylo.blogspot.com Home page: http://taxonomy.zoology.gla.ac.uk/rod/rod.html
tdwg-content mailing list tdwg-content@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-content
-- Dr. Donat Agosti Research Associate, American Museum of Natural History and Smithsonian Institution
Email: agosti@amnh.org Web: http://antbase.org CV: http://research.amnh.org/entomology/social_insects/agosticv_2003.html
Swiss Residence Elahieh Ave. Khazer no. 74 19649 Teheran Iran
+98-21-2200 8765 (office) +98-21-2260 6160 (home) +98-919-489 2744 (mobile) +1-202-558 0330 (skype-in US) +41-44-5862911 (skype-in Switzerland)
-- Dr. Donat Agosti Research Associate, American Museum of Natural History and Smithsonian Institution
Email: agosti@amnh.org Web: http://antbase.org CV: http://research.amnh.org/entomology/social_insects/agosticv_2003.html
Swiss Residence Elahieh Ave. Khazer no. 74 19649 Teheran Iran
+98-21-2200 8765 (office) +98-21-2260 6160 (home) +98-919-489 2744 (mobile) +1-202-558 0330 (skype-in US) +41-44-5862911 (skype-in Switzerland)
tdwg-content mailing list tdwg-content@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-content
--
Pete DeVries Department of Entomology University of Wisconsin - Madison 445 Russell Laboratories 1630 Linden Drive Madison, WI 53706 TaxonConcept Knowledge Base http://www.taxonconcept.org/ / GeoSpecies Knowledge Base http://lod.geospecies.org/ About the GeoSpecies Knowledge Base http://about.geospecies.org/
tdwg-content mailing list tdwg-content@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-content