[go: nahoru, domu]
Skip to content

Latest commit

 

History

History
12 lines (8 loc) · 449 Bytes

README.md

File metadata and controls

12 lines (8 loc) · 449 Bytes
Raw

CRF-Chucking

Using Conditional Random Fields to Chunk the words in a sentence

Algorithm: Gradient descent using l2 regularization

Features used:

  • Current Word, Previous Word, Previous to Previous Word (word vectors). 50 dimensional Glove vectors.
  • Current POS, Previous POS, Previous to Previous POS.
  • Previous chunk labels, previous to the current position.
  • Morphological features: word stem and word affixes of current word