5. POS Tagging

Question 1

What is POS Tagging?

Answer 1

A Part-Of-Speech Tagger reads text in some language and assigns parts of speech to each word (and other token), such as noun, verb, adjective, etc.

eg.
I want a ticket:
I -> pronoun
want -> verb
a -> det
ticket -> noun

Question 2

What are Universal POS Tags? Give a few examples of tags.

Answer 2

Universal POS tags are a list of part-of-speech tags (POS tags for short), i.e. labels used to indicate the part of speech and sometimes also other grammatical categories (case, tense etc.) of each token in a text corpus.

Some tags:
- ADJ (adjective), ADV (adverb), NOUN (noun), VERB (verb), PUNCT (punctuation) and others.

Question 3

What is the naive approach to perform POS Tagging? How does it perform?

Answer 3

Naive Approach:
- Assigning each word its most frequent POS tag, and assigning all unknown words the tag NOUN.

Performs surprisingly well with around 90% accuracy, but there are exception where it does not perform as well.

Question 4

Create the tables for tag given previews tag and word given tag for POS Tagging given the following corpus:

• John/PROPN is/VERB expected/VERB to/PART race/VERB
• This/DET is/VERB the/DET race/NOUN I/PRON wanted/VERB
• Bring/VERB this/DET to/PART the/DET race/NOUN

Answer 4

Check Lecture Notes 5. POS Tagging slides 17 and 18.

Question 5

What is the Viterbi algorithm? What’s its complexity?

Answer 5

The Viterbi algorithm is a dynamic programming algorithm for finding the most likely sequence of hidden states—called the Viterbi path—that results in a sequence of observed events, especially in the context of Markov information sources and hidden Markov models.

Complexity: O(SN^2) where S is the length of the input and N is the number of states in the model.

Question 6

What is an alternative to Hidden Markov Models (HMM)?

Answer 6

An alternative is a sequence version of logistic regression classifier, the Maximum Entropy Classifier (MEMM), a discriminative model to directly estimate posterior.

Question 7

Compare HMM and MEMM.

Answer 7

Maybe read this, Luscia’s explanation is disgustingly bad

https://medium.com/@Alibaba_Cloud/hmm-memm-and-crf-a-comparative-analysis-of-statistical-modeling-methods-49fc32a73586