Improving Part-Of-Speech Tagging Accuracy for Croatian by Morphological Analysis (Report‪)‬

1 Introduction After obtaining satisfactory results of the preliminary experiment with applying a second order hidden Markov model part-of-speech/morphosyntactic tagging paradigm by using TnT tagger on Croatian texts, we decided to attempt reaching a higher level of accuracy based on these results. Detailed description of the previous experiment is given in (Agid and Tadid 2006) and TnT tagger is described in (Brants 2000). Please note that abbreviation HMM is used instead hidden Markov model and PoS (MSD) tagging instead part-of-speech (morphosyntactic) tagging further in the text.