... (76,739,723 sentences) from the MEDLINE abstracts tokenised by McIn-tosh and Curran (2008). These sentences werePOS-tagged and parsed twice, once as for thenewswire and Wikipedia data, and then again, ... adapted to suit the parser by train-ing on the combination of these sets and the stan-dard training corpora. We applied standard evalu-ation metrics for speed and accuracy, and exploredthe source ... Budapest, Hungary.John D. Lafferty, Andrew McCallum, and FernandoC. N. Pereira. 2001. Conditional random fields:Probabilistic models for segmenting and labeling se-quence data. In Proceedings of...