Parts-of-Speech Tagging (POS) using Hidden Markov Model (HMM )and (MEMM)

Last updated on May 11, 2021

Parts of speech tagging the process of assigning a part-of-speech tag (Noun, Verb, Adjective…) to each word in an input text. I have trained two different model HMM which is generative model and MEMM which is discriminative.

  • Train a Hidden Markov Model(HMM) and Bigram Maximum Entropy Model for POS tagging

  • Hidden Markov Model(HMM)

    • Compute transition matrix and emission matrix
    • Compute viterbi algorithm and evaluate accuracy of model on Wall Street Journal (WSJ)

  • Maximum Entropy Model

    • Preprocess and create features (eg. wordshape,number,hyphen) for each word in sentence
    • Train a logistic regression and compute log probability distributions over the tagset.
    • Compute viterbi algorithm and evaluate accuracy of model on Brown corpus
Machine Learning NLP

Related