[Taxacom] taxonomic names databases

[godfoder]

This is a great example of identifier cross linking in action and 
working pretty well. I think that, even if you disagree with GBIF's 
choices in constructing the backbone, that this level of cross linking 
among disparate taxonomies is incredibly valuable. It would be great if 
we could search every database for any identifier of our choosing and 
expect a result such as your proposed searching of GBIF and OTT based on 
IRMNG IDs. I think GBIF is working to enable that at least to some 
extent, although the API call I expected to do this did not work: 
http://api.gbif.org/v1/species?datasetKey=0938172b-2086-439c-a1dd-c21cb0109ed5&sourceid=10561410 
. For now though, we're still a ways away from that goal.

Although this is liable to get my head bitten off, I think that this 
cross linking makes GBIF's taxon identifiers much more valuable than 
IRMNG's even if they are potentially less correct (hang on, don't lynch 
me just yet) scientifically. The (oversimplified) theory behind Google's 
famous page rank algorithm is that a node on a graph can be ranked based 
on the number of inbound and outbound links. This turns out to be a 
fantastic basis for ranking the relevance of a huge amount of different 
types of information. GBIF has done the work to build out the outbound 
links from their taxonomic identifiers, as well as building a whole 
internal link network between taxonomy and occurrence data. For projects 
with limited resources looking to inter-relate data this means that 
whether you're a data aggregator, species modeling project , working on 
a competing backbone, or focused on a specific taxonomic niche knowing 
the GBIF taxonIDs of your subjects is very useful.

This doesn't mean I think that other taxonomies and projects should 
abandon their own identifiers, quite to the contrary. I think that local 
identifiers often capture a level of specificity that is virtually 
impossible to represent in aggregated data sets. What I do think, is 
that projects like IRMNG who are looking to capture value for 
themselves, should actively seek out and integrate these high value 
identifiers into their own databases. If both Index Fungorum and IRMNG 
did this, it would be possible to link the two databases based on their 
shared knowledge of GBIF IDs. It would also make it much easier in the 
future to do a cross walk between the two database to directly 
incorporate each-others identifiers since you could start from that 
shared knowledge.

- Alex

On 08/31/2016 03:18 PM, Tony Rees wrote:
> Thinking a step further (again, possibly only talking to myself here
> :) ), even some of the above "cross linking" between datasets could
> potentially be outsourced, further spreading that load - at least some
> of the building blocks are there (but not the IRMNG <> ION mapping
> anywhere else at this time, so far as I can see).
>
> For example, IRMNG data dumps are already used (as one component) in
> "taxonomic backbone" building exercise by GBIF, Open Tree of Life,
> Encyclopedia of Life, Atlas of Living Australia and possibly others,
> and this process will/should be refreshed with new releases of IRMNG
> and other data sources. To do this, such "name aggregators" will
> already be comparing names and identifiers from IRMNG to those from
> other sources and most likely, keeping whatever mappings they make
> "somewhere", maybe even exposing them, as here for the previously used
> example (species name in Diptera):
>
> Open Tree: https://tree.opentreeoflife.org/opentree/argus/ottol@4474544/Aagaardia-protensa
> GBIF: http://www.gbif.org/species/1446570
> (the relevant page on EOL does not appear to have mapped the IRMNG
> entry, but no matter).
>
> Now, so far as I know there is not a publicly advertised way to search
> Open Tree, GBIF or EOL by IRMNG ID, but from purely a technical
> viewpoint, there is no reason why there could not be. For example
> taking the Open Tree page, that then already maps IRMNG: 10561410 to
> both OTT: 4474544 and GBIF: 1446570, and the GBIF page maps the IRMNG
> entry to the same one in GBIF (1446570) and also to 4 other data
> databases ("checklists" in GBIF speak) plus whatever occurrence data
> GBIF may hold.
>
> The upshot of the above is that if one could search the above systems
> by IRMNG identifier, with certain caveats (such as the relevant cross
> mapping being correctly done) IRMNG what not need to know what
> identifiers Open Tree, GBIF etc. were using for the same names (these
> could be retrieved on the fly or even not needed) and in addition,
> could leverage whatever additional links such systems have already
> made.
>
> For example, in my previous post I said it would be nice for a user
> (human or machine) to be able to go from fungal records in IRMNG to
> the relevant entry in Index (or Species) Fungorum, which might for
> example contain the original literature for a particular taxon name
> and save IRMNG the task of compiling that. It turns out that this
> should already be possible : for example the Open Tree page for the
> fungus Amanita caesarea already lists the following:
>
> Reference taxonomy:
> OTT: 769441
>
> Source taxonomy:
> Index Fungorum: 208468
> NCBI: 67695
> GBIF: 5452695
> GBIF: 5240269
> IRMNG: 10381934
>
> This means they have already cross mapped IRMNG: 10381934 to Index
> Fungorum: 208468, saving me the trouble, and there *should* be a way
> to provide users with a route to go from the (presently bare bones)
> IRMNG page here:
> http://www.irmng.org/aphia.php?p=taxdetails&id=10381934 to the
> equivalent in IF:
> http://www.indexfungorum.org/Names/namesrecord.asp?RecordID=208468
>
> Which contains the original publication information "Citations in
> published lists or literature: Saccardo's Syll. fung. V: 8; XII: 905;
> XIX: 44" as well as the magic additional link "Page Image for
> Protologue:" which points here:
> http://www.librifungorum.org/Image.asp?ItemID=82&ImageFileName=0252b.jpg
>
> Now that's looking like a win (thanks, Paul and the Open Tree folk in
> this instance).
>
> Comments, anyone?
>
> Regards - Tony
>
>