RAMBO 800+: A Corpus for the Development of Gene/Protein Recognition from Rare and Ambiguous Abbreviations

We release the RAMBO 800+ corpus providing manual annotations for Rare and AMBiguOus abbreviations of gene names in about 800 MEDLINE abstracts. It can be used to train gene recognition systems for this class of abbreviations, as discussed in Hartung et al. (BioNLP 2014). The corpus covers eight gene name abbreviation types: AHR, CLI, CLU, COPD, HF, MOX, PLS, SAH. For each of these types, 100 (in case of MOX: 81) abstracts have been randomly sampled from MEDLINE. In each of these abstracts, every mention of an abbreviation of interest has been manually annotated as denoting a gene/protein or not. Plus, all other tokens in the 800 abstracts have been annotated in the same way.