Why the ICZN is in trouble

There are many reasons why the International Commission on Zoological Nomenclature (ICZN) is in trouble, but fundamentally I think it's because of situation illustrated by following diagram.


Based on an analysis of the Index of Organism Names (ION) database that I'm currently working on, there are around 3.8 million animal names (I define "animal" loosely, the ICZN covers a number of eukaryote groups), of which around 1.5 million are "original combinations", that is, the name as originally published. The other 2 million plus names are synonyms, spelling variations, etc.

Of these 3.8 million names the ICZN itself can say very little. It has placed some 12,600 names (around 0.3% of the total) on its Official Lists and Indexes (which is where it records decisions on nomenclature), and its new register of names, ZooBank, has less than 100,000 names (i.e., less than 3% of all animal names).

The ICZN doesn't have a comprehensive database of animal names, so it can't answer the most basic questions one might have about names (e.g., "is this a name?", "can I use this name, or has somebody already used it?", "what other names have people used for this taxon?", "where was this name originally published?", "can I see the original description?", "who first said these two names are synonyms?", and so on). The ICZN has no answer to these questions. In the absence of these services, it is reduced to making decisions about a tiny fraction of the names that are in use (and there is no database of these decisions). It is no wonder that it is in such trouble.

Nomenclator Zoologicus meets Biodiversity Heritage Library: linking names directly to literature

Following on from my previous post on microcitations I've blasted all the citations in Nomenclator Zoologicus through my microcitation service and created a simple web site where these results can be browsed.

The web site is here: http://iphylo.org/~rpage/nz/.

To create it I've taken a file dump of Nomenclator Zoologicus provided by Dave Remsen and run all the citations through the microcitation service, storing the results in a simple database. You can search by genus name, author and year, or publication. The search is pretty crude, and in the case of publications can be a bit hit and miss. Citations in Nomenclator Zoologicus are stored as strings, so I've used some crude rules to try and extract the publication name from the rest of the details (such as page numbering).

To get started, you can look at names published by published by Distant in 1910, which you can see below:



If the citation has been found you can click on the icon to view the page in a popup, like this:



You can also click on the page number to be taken to that page in BHL.


I've also added some other links, such as to the name in the Index to Organism Names, as well as bibliographic identifiers such as DOIs, Handles, and links to JSTOR and CiNii.

So far only 10% of Nomenclator Zoologicus records have a match in BHL, which is slightly depressing. Browsing through there are some obvious gaps where my parser clearly failed, typically where multiple pages are included in the citation, or the citation has some additional comments. These could be fixed. There are also cases where the OCR text is so mangled that a match has been rejected because the genus name and text were too different.

This has been hastily assembled, but it's one vision of a simple service where we can go from genus name to being able to see the original publication of that name. There are other things we could do with this mapping, such as enabling BHL to tell users that the reference they are looking at is the original source of a particular name, and enabling services that use BHL content (such as EOL and Atlas of Living Australia to flag which reference in BHL is the one that matters in terms of nomenclature.

BHL and OCR

Some quick notes on OCR. Revisiting my DjVu viewer experiments it really struck me how "dirty" the OCR text is. It's readable, but if we were to display the OCR text rather than the images, it would be a little offputting. For example, in the paper A new fat little frog (Leptodactylidae: Eleutherodactylus) from lofty Andean grasslands of southern Ecuador (http://biostor.org/reference/229) there are 15 different variations of the frog genus Eleutherodactylus:

• Eleutherodactylus
• Eleutheroclactylus
• Eleuthewdactyliis
• Eleiitherodactylus
• Eleuthewdactylus
• Eleuthewdactylus
• Eleutherodactyliis
• Eleutherockictylus
• Eleutlierodactylus
• Eleuthewdactyhts
• Eleiithewdactylus
• Eleutherodactyhis
• Eleiithemdactylus
• Eleuthemdactylus
• Eleuthewdactyhis

Of course, this is a recognised problem. Wei et al. Name Matters: Taxonomic Name Recognition (TNR) in Biodiversity Heritage Library (BHL) (hdl:2142/14919) found that 35% of names in BHL OCR contained at least one wrong character. They compared the performance of two taxonomic name finding tools on BHL OCR (uBio's taxonFinder and FAT), neither of which did terribly well. Wei et al. found that different page types can influence the success of these algorithms, and suggested that automatically classifying pages into different categories would improve performance.

Personally, it seems to me that this is not the way forward. It's pretty obvious looking at the versions of "Eleutherodactylus" above that there are recognisable patterns in the OCR errors (e.g., "u" becoming "ii", "ro" becoming "w", etc.). After reading Peter Norvig's elegant little essay How to Write a Spelling Corrector, I suspect the way to improve the finding of taxonomic names is to build a "spelling corrector" for names. Central to this would be building a probabilistic model of the different OCR errors (such as "u" → "ii"), and use that to create a set of candidate taxonomic names the OCR string might actually be (the equivalent of Google's "did you mean", which is the subject of Norvig's essay). I had hoped to avoid doing this by using an existing tool, such as Tony Rees' TAXAMATCH, but it's a website not a service, and it is just too slow.

I've started doing some background reading on the topic of spelling correction and OCR, and I've created a group on Mendeley called OCR - Optical Character Recognition to bring these papers together. I'm also fussing with some simple code to find misspellings of a given taxonomic names in BHL text, use the Needleman–Wunsch sequence alignment algorithm to align those misspellings to the correct name, and then extract the various OCR errors, building a matrix of the probabilities of the various transformations of the original text into OCR text.

One use for this spelling correction would be in an interactive BHL viewer. In addition to showing the taxonomic names that uBio's taxonFinder has located in the text, we could flag strings that could be misspelt taxonomic names (such as "Eleutherockictylus") and provide an easy way for the user to either accept or reject that name. If we are going to invite people to help clean up BHL text, it would be nice to provide hints as to what the correct answer might be.

Are names really the key to the big new biology?

David ("Paddy") Patterson, Jerry Cooper, Paul Kirk, Rich Pyle, and David Remsen have published an article in TREE entitled "Names are key to the big new biology" (doi:10.1016/j.tree.2010.09.004). The abstract states:

Those who seek answers to big, broad questions about biology, especially questions emphasizing the organism (taxonomy, evolution and ecology), will soon benefit from an emerging names-based infrastructure. It will draw on the almost universal association of organism names with biological information to index and interconnect information distributed across the Internet. The result will be a virtual data commons, expanding as further data are shared, allowing biology to become more of a ‘big science’. Informatics devices will exploit this ‘big new biology’, revitalizing comparative biology with a broad perspective to reveal previously inaccessible trends and discontinuities, so helping us to reveal unfamiliar biological truths. Here, we review the first components of this freely available, participatory and semantic Global Names Architecture.

Do we need names?

Reading this (full disclosure, I was a reviewer) I can't wondering whether the assumption that names are key really needs to be challenged. Roger Hyam has argued that we should be calling time on biological nomenclature, and I wonder whether for a generation of biologists brought up on DNA barcodes and GPS, taxonomy and names will seem horribly quaint. For a start, sequences and GPS coordinates are computable, we can stick them in computers and do useful things with them. DNA barcodes can be used to infer identity, evolutionary relationships, and dates of divergence. Taken in aggregate we can infer ecological relationships (such as diet, e.g., doi:10.1371/journal.pone.0000831), biogeographic history, gene flow, etc. While barcodes can tells us something about an organism, names don't. Even if we have the taxonomic description we can't do much with it — extracting information from taxonomic descriptions is hard.

Furthermore, formal taxonomic names don't seem terribly necessary in order to do a lot of science. Patterson et al. note that taxa may have "surrogate" names":

Surrogates include provisional names and specimen, culture or strain numbers which refer to a taxon. 'SAR-11' ('SAR' refers to the Sargasso Sea) was a surrogate name given in 1990 to an important member of the marine plankton. Only a decade later did it become known as Pelagibacter ubique.

The name Pelagibacter ubique was published in 2002 (doi:10.1038/nature00917), although as a Candidatus name (doi:10.1099/00207713-45-1-186), not a name conforming to the International Code of Nomenclature of Bacteria. I doubt the lack of a name that follows this code is hindering the study of this organism, and researchers seem happy to continue to use 'SAR11'.

So, I think that as we go forward we are going to find nomenclature struggling to establish its relevance in the age of digital biology.

If we do need them, how do we manage them?
If we grant Patterson et al. their premise that names matter (and for a lot of the legacy literature they will), then how do we manage them? In many ways the "Names are key to the big new biology" paper is really a pitch for the Global Names Architecture or GNA (and it's components GNI, GNITE, and GNUB). So, we're off into alphabet soup again (sigh). The more I think about this the more I want something very simple.

Names
All I want here is a database of name strings and tools to find them in documents. In other words, uBio.

Documents
Broadly defined to include articles, books, DNA sequences, specimens, etc. I want an database of [name,document] pairs (BHL has a huge one), and a database of documents.



Realistically, given the number and type of documents there will be several "document" databases, such as GenBank and GBIF. For citations Mendeley looks very promising. If we had every taxonomic publication in Mendeley, tagged with scientific names, then we'd have the bibliography of life. Taxonomic nomenclators would be essentially out of business, given that their function is to store the first publication of a name. Given a complete bibliography we just create a timeline of usage for a name and note the earliest [name,document] pair:

Taxonomy
There are a few wrinkles to deal with. Firstly, names may have synonyms, lexical variants, etc. (the Patterson et al. paper has a nice example of this). Leaving aside lexical variants, what we want is a "view" of the [name,document] pairs that says this subset refer to the same thing (the "taxon concept").



We can obsess with details in individual cases, but at web-scale there are only two ones that spring to mind. The first is the Catalogue of Life, the second is NCBI. The Catalogue of Life lists sets of names and reference that it regards as being the same thing, although it does unspeakable things to many of the references. In the case of NCBI the "concepts" would be the sets of DNA sequences and associated publications linked to the same taxonomy id. Whatever you think of the NCBI taxonomy, it is at least computable, in the sense that you could take a taxon and generate a list of publications 'about" that taxon.

So, we have names, [name,document] pairs, and sets of [name,document] pairs. Simples.