๐Ÿ•ถ Brill tagger

= a transformation based POS-tagger


For each word in corpus:

  1. Tag the word with the most likely POS
  2. Choose a transformation that deterministically replaces an existing tag with a new one to improve the error rate
  3. Apply the transformation

Example transformations

1NNVBPrevious tag is TOto/TO race/NN โ†’ VB
2VBPVBOne of the previous tags is MDmight/MD vanish/VBP โ†’ VB