839 terms were extracted from the FNA glossary, compared to 1080 terms in the plant anatomical entity branch of the PO. Using text matching, Obol mapped 264 FNA terms to 313 existing PO terms or synonyms, including 49 FNA terms that matched more than one PO term or synonym. Most duplicate matches arose because the PO has many synonyms in Spanish that are identical to the English term name. Only 30 Obol matches had to be removed, in cases NSF-IOS: 0822201 to the Plant Ontology Project, and the Flora of North America Association.
Traditional taxonomic literature can provide a wealth of data, but access to that data is limited by its free-text format. Taxonomic treatments such as the Flora of North America (FNA Editorial Committee 1993) consist of terse descriptions of the characters used to identify taxa, such as: “…Leaves usually alternate or opposite, sometimes in basal rosettes, rarely in whorls; rarely stipulate, usually petiolate, sometimes sessile…” Converting taxonomic descriptions to computer-readable format makes them available for automatic retrieval and large-scale analyses. Ontologies such as the Plant Ontology (PO) play a central role in automatic annotation, by providing semantic meaning for the words in a description. We used automated and manual methods to map terms from the Categorical Glossary for the Flora of North America Project (http://128.2.21.109/fmi/xsl/FNA/home.xsl) to the PO.
Terms from the pre-existing categories of “structure”, “feature”, or “nominative” were extracted from the FNA glossary, roughly corresponding to the PO class plant anatomical entity. An automated mapping to PO release 16 was done using Obol software (Mungall 2004). We manually checked the automated mapping, and removed any matches that were incorrect. Remaining glossary terms were either manually mapped to existing PO terms, classified as inappropriate for the PO, or marked to be added to the PO.