Perbandingan Algoritma Conditional Random Field dan Hidden Markov Model pada Pos Tagging Bahasa Indonesia

Abstract

Twitter is now an alternative source of real time
information for the public. Technological developments
cover all aspects of life, one of which is the field of language.
Natural Language Processing (NLP) devices developed to
support those needs are POS Tagger. This research use 10
tweets and HMM algorithm get 62,7% accuracy level while
Conditional Random Field algorithm get 71%. This shows
that CRF is better for performing POS tagging in Indonesian
on Twitter. HMM and CRF can handle tagging of words that
are not in the corpus but the results are not very good.