In natural language processing, Part-of-Speech (POS) tagging refers to the process of assigning each word (or nonword token) in text with a tag identifying its part of speech, drawn from some fixed set of tags. A number of different tagsets are in use; one of the most frequently applied is the Penn Treebank tagset, which contains 36 POS tags and 12 punctuation and other tags (Marcus et al. 1993). POS tagging is often approached as a sequential labeling task addressed with machine learning methods such as Hidden Markov Models and Conditional Random Fields (Manning and Schütze 1999). For training accurate POS taggers for biomedical domain texts, domain corpora manually annotated for POS tags such as GENIA corpus are typically applied. The parts of speech of words provide useful information for a number of tasks ranging from word sense disambiguation to named entity recognition and information extractionand POS tagging is a frequently applied...