hunpos follows the architecture of Thorsten Brandt's TnT system (tag n-gram HMM, with the suffix guessing emission model for unseen words), but with the ability to incorporate the output of a morphological analyzer, using the output of the morphological analyzer to constrain suffix guessing when meeting unseen words. As an improved, open source reimplementation of TnT, it is frequently used for the task of POS-tagging and morphological disambiguation, and is one of the standard building blocks and baselines when creating new POS-tagging systems and evaluating their precision.

You don’t have the permission to edit this resource.

People who looked at this resource also viewed the following: