Part-of-Speech (PoS) Tagging

PoS Tagging - Methods

Rule-Based Tagging

  • given input sentence we assign each word ALL POSSIBLE PoS tags, according to dictionary
  • then we use rules to REMOVE tags, until each word has exactly one PoS Tag
  • rules a hand built (not auto-learned)

Transformation-Based PoS Tagging

  • given input sentence we assign each word a SINGLE PoS tag, according to dictionary
  • then we use rules to TRANSFORM tags, until no more transformations are made
  • rules can be auto-learned by feeding it training corpus

Statistical-Based Tagging

  • given input sentence we assign a PoS Tag to each word that maximizes the probability of being “correct”
  • probabilities can be auto-learned by feeding it training corpus