The following changes were made to the Treebank POS tag-set (0) The tag UH is used for speech pauses/interjections like okay, UH, well,... (1) The tag /TO is now restricted to infinitival to. When to is used as a preposition it will be tagged /IN like all other prepositions. I want to/TO go. I went to/IN the store. (2) To eliminate the ambiguity of 's/VBZ which can stand for either 'is' or 'has', we now use two new tags: /BES for 's when it is contracted from is and /HVS when it is contracted from has. He 's/BES a big boy now. She's/HVS got lots of money. (3) A new tag, /XX is used for partial words, but only when it isn't clear from the context what the word is. If it is clear what the word is, it is tagged as usual. [ we/PRP ] got/VBD [ one/CD ] [ -s/XX ] ,/, [ one/CD cut/VBN out/RP ] on/IN [ the/DT table/NN ] saw/NN ,/, (4) A new tag, /GW is used to join morphs that act as a single `word' but have been written with spaces. These are mostly inflected ``words'' of the type 'B S ing', 'R V -er', 'M and M s', etc., but also things like T V, where the T has no part-of-speech. /GW stands for 'goes with' and is attached to any word/morph which forms a 'word' with what follows. The POS of the whole compound is then marked on the last element. R/GW V/GW -er/NN B/GW S/GW ing/VBG T/GW V/NN