Taxonomic papers are generally a synthesis of a limited set of elements, including text descriptions, scientific names and nomenclatural acts, literature references, images, specimen occurrence records, and increasingly DNA sequences. The role of the author is to link specimens (with their associated occurrence records) to nomenclature, express observations and hypotheses as text, and document observations with images and quantities. In traditional taxonomic publishing, all of these elements are merged together into a document. But what if these elements could simultaneously be released and maintained as discrete data tied to the publication? This would have effects both within and beyond the taxonomic community. With access to data elements in electronic form, data consumers (for example, taxonomists, ecologists, conservationists, molecular biologists) could use the publication in more flexible ways. This could aid in the taxonomic utility of the work, facilitate the recognition of new discoveries, and increase the testability (and hence the scientific quality) of the work. The key to unlocking this potential is semantic tagging.

Semantic tagging is a method of assigning markers, or tags, to a text string so the meaning of that string is discoverable and readable by computers [5]. Data elements organized and tagged according to accepted standards are trivial to combine. By contrast, reconciling the content of multiple traditional publications on the same group can require dedicated study and effort. Once in parsed form, data are available for recombination and repurposing. Associating data elements with a publication adds a measure of credibility based on the reputation of the authors, the review process of the publication venue, and the date of publication [6]. This contrasts with, for example, the museum-collections-based data aggregation model that currently dominates GBIF (Global Biodiversity Information Facility, http://data.gbif.org), where data credibility rests with the originating institution. To the extent that the corpus of taxonomic literature can be parsed and aggregated by various cybertaxonomic repositories, those repositories become powerful tools for fundamental information about the state of biodiversity knowledge, meta-analysis, and data reuse for a wide range of applications, including public outreach [1, 3, 4, 7, 8].

Once we agree that it would be desirable to have semantically tagged taxonomic data elements digitally interlinked and associated with publications, there are several ways to approach content development. The strategy we focus on is the journal-centered approach, considering both retrospective and prospective content. The retrospective portion involves converting legacy publications from their current formats (for example, print, PDFs) into parsed and digitally distributed content. The prospective portion involves semantic tagging embedded during the editorial production process. It seems unrealistic for the large number of individual taxonomists, each publishing a handful of taxonomic papers per year, to keep current with the technology. But publishers of journals with an emphasis in taxonomy are better positioned to develop and maintain efficient and current processes.

The infrastructure is already in place for cybertaxonomic resources to recognize and aggregate content appropriate to their focus from documents semantically tagged according to XML standards [5]. We see a future where major taxonomic journals routinely expose new and legacy content in ways that can be discovered, aggregated, and distributed by a community of cybertaxonomic repositories (Figure 1). The source XML documents might be hosted by individual journals or kept in a common repository such as Plazi (http://plazi.org/). To facilitate the integration of biological information across diverse sources, each digitized publication should be distinguished by a globally unique identifier (GUID), such as a registered DOI (digital object identifier) or LSID (Life Science Identifier), linking the data elements back to the original source, author, and journal [7].

The undisputed leader in prospective parsing and dissemination of taxonomic content is Pensoft (http://www.pensoft.net/), publisher of cutting edge open access cybertaxonomy journals, including ZooKeys, PhytoKeys, and MycoKeys. The first of these, ZooKeys, started with a revolutionary publishing model designed to disseminate biodiversity data in both traditional and innovative ways. This included registering all new nomenclatural acts with the ZooBank database (http://www.zoobank.org/) and providing species descriptions to the Encyclopedia of Life (http://www.eol.org/) as a routine part of their work flow. Pensoft adopted Plazi's TaxPub XML schema as a starting point [5]. Over time, ZooKeys and her sister journals continued to push the envelope of cybertaxonomic publishing, constantly looking for new avenues for sharing biodiversity data. It has become clear that parsing content into semantically tagged elements and publishing primary data in standardized digital form are an efficient and powerful combination for repurposing content from primary source taxonomic literature [5, 8]. Open access is essential to facilitate the flow of data from taxonomic literature to cybertaxonomic repositories. The many benefits of open access in taxonomic publishing have been convincingly enumerated elsewhere [9, 10]. They certainly include improving accessibility of publications internationally and beyond the immediate taxonomic community.

We would encourage more taxonomic journals to adopt some of Pensoft's routine practices. For prospective publishing, there are similarities between document layout and XML markup that might be combined to mitigate any added burden. However, the costs and benefits of this approach will depend largely on the volume of taxonomic papers published by a particular journal or publisher. Based on experience with Plazi and Pensoft, XML markup multiplies production costs by 5 and takes 0.5 to 2 minutes per page. It may not make sense for journals to provide XML markup if covering a range of topics that only occasionally includes taxonomy. In such cases, responsibility for XML markup may be better placed with the author. A new generation of tools is starting to appear that allow authors to semanticize content while writing a manuscript. These include the Publication Module in EDIT's Scratchpads (http://scratchpads.eu/) [11] and the Pensoft Writing Tool (PWT; http://www.pensoft.net/services-for-journals).

Although thousands of journals have published taxonomic papers, just a few are responsible for the vast majority. Focusing resources on marking up the legacy content of major taxonomic journals is an efficient way to move a large body of content online. Addressing the long tail of taxonomic content in smaller journals will require a more distributed approach (Figure 2).

The GoldenGATE XML Markup Editor (free download from http://plazi.org/?q=GoldenGATE) is an innovative and powerful tool for converting taxonomic papers into XML documents. GoldenGATE analyses the PDF of a taxonomic paper and automatically recognizes the semantic meaning of the various elements. A human operator then checks and refines the markup. Because journals tend to follow particular formatting conventions, GoldenGATE can be optimized for accuracy with a particular journal - one reason to focus effort by journal, rather than by taxon for example. The result is an XML document based on the TaxonX schema (http://www.taxonx.org/). This document can then be made available to cybertaxonomic data aggregators. In its current form, GoldenGATE requires study, training (estimated 2 to 3 days), and practice before an operator can be considered proficient. As the value of marked up legacy content from major journals becomes more widely appreciated, developers will be incentivized to create a new generation of software tools tailored for a more distributed user community. Taxonomists will start filling in the gaps by marking up and sharing articles relevant to their own work.

Taxonomic materials that have not been published in open access journals are not excluded from retrospective semantic markup [12]. Under a limited set of circumstances, 'Fair Use' clauses of copyright law permit copying protected materials without consulting the copyright holder (but see [3]). However, data mining of copyrighted materials remains controversial and is opposed by some publishers [13].

The journal-centric approach to cybertaxonomy should be understood to mean that we are not advocating any change in the way taxonomists conduct their science. Rather, taxonomists can focus effort and time on their research and on submitting manuscripts to journals that provide traditional peer review and technical editing, but now also make the information available to the community of online cybertaxonomic resources. This will enfranchise new classes of users who can use the fruits of taxonomic research in ways anticipated and unanticipated by the authors. Barlow's economic paradigm teaches us that the more accessible, comprehensive and usable taxonomic information is, the more it will be valued. Therefore, the goal of taxonomy should not simply be to describe every species on Earth, but to make that information accessible.

Understanding biodiversity is such a massive challenge that the traditional approach of working in relative isolation to produce paper publications for colleagues and libraries is simply too inefficient. But there is an alternative model: increase the value of knowledge by giving it away to the world on a massive scale, and empower others to help it grow. More than 250 years after taxonomists began describing and cataloging species, the internet and an ethic of information sharing let us organize biodiversity information in a way that is responsive to the questions and the challenges of our changing world.

In May to June 2011, the authors assembled at the Encyclopedia of Life's Biodiversity Synthesis Center (BioSynC) in Chicago's Field Museum of Natural History. Our objective was to formulate practical tools and strategies for the efficient transfer of taxonomic data from legacy and contemporary paper publications to major cybertaxonomic repositories. We are grateful to BioSynC for their generous support of our meeting. In particular, we thank Mark Westneat and Audrey Aronowsky for being such gracious hosts. GoldenGATE was developed with support from the Karlsruhe Institute of Technology (KIT). Thanks to Naturalis librarian Tom Gilissen for help with the Zoological Records search.