Although the Web has transformed science publishing, scientific papers themselves are still essentially "black boxes", with much of their content intended for human readers only. Typically, computer-readable metadata associated with an article is limited to bibliographic details. By expanding article metadata to include taxonomic names, identifiers for cited material (e.g., publications, sequences, specimens, and other data), and geographical coordinates, publishers could greatly increase the scientific value of their digital content. At the same time this will provide novel ways for users to discover and navigate through this content, beyond the relatively limited linkage provided by bibliographic citation.
As a proof of concept, my entry in the Elsevier Grand Challenge extracted extended metadata from a set of articles from the journal Molecular Phylogeny and Evolution and used it to populate an entityattribute-value database. A simple web interface to this database enables an enhanced display of the content of an article, including a map of localities mentioned either explicitly or implicitly (through links to geotagged data), taxonomic coverage, and both data and citation links. Metadata extraction was limited to information listed in tables in the articles, such as GenBank sequences and specimen codes. The body of the article was not used, a restriction that was deliberate to demonstrate that making extended metadata available does not require a journal's publisher to make the full-text freely available (although this is desirable for other reasons).