Virtual Labs

Building POS Tagger

What is the primary purpose of Part-of-Speech (POS) tagging in Natural Language Processing?

a: To assign grammatical categories to each word in a sentence Explanation

Explanation

b: To translate text from one language to another Explanation

Explanation

c: To count the frequency of words in a document Explanation

Explanation

d: To generate new sentences automatically Explanation

Explanation

Which of the following words can have multiple POS tags depending on context?

a: Book Explanation

Explanation

b: The Explanation

Explanation

c: Always Explanation

Explanation

d: Blue Explanation

Explanation

In this experiment, which algorithms are used for POS tagging?

a: Hidden Markov Model (HMM) only Explanation

Explanation

b: Conditional Random Field (CRF) only Explanation

Explanation

c: Both Hidden Markov Model (HMM) and Conditional Random Field (CRF) Explanation

Explanation

d: Deep Neural Networks only Explanation

Explanation

What generally happens to POS tagging accuracy when you increase the training corpus size?

a: Accuracy typically improves up to a certain point Explanation

Explanation

b: Accuracy always decreases Explanation

Explanation

c: Accuracy remains exactly the same Explanation

Explanation

d: Accuracy changes randomly Explanation

Explanation

Which context feature typically provides the best balance of accuracy and computational efficiency for POS tagging?

a: Unigram (no context) Explanation

Explanation

b: Bigram (previous word/tag) Explanation

Explanation

c: Trigram (two previous words/tags) Explanation

Explanation

d: 5-gram (four previous words/tags) Explanation

Explanation

What is a key advantage of Conditional Random Fields (CRF) over Hidden Markov Models (HMM) for POS tagging?

a: CRF can incorporate a wider range of features and dependencies Explanation

Explanation

b: CRF is always faster than HMM Explanation

Explanation

c: CRF requires less training data Explanation

Explanation

d: CRF only works with small vocabularies Explanation

Explanation

Why might POS tagging accuracy differ between English and Hindi in this experiment?

a: Different scripts, morphological complexity, and word order patterns Explanation

Explanation

b: Hindi is inherently more difficult to process Explanation

Explanation

c: English has fewer grammatical categories Explanation

Explanation

d: The algorithms only work well for English Explanation

Explanation

In the Viterbi algorithm used with HMM for POS tagging, what is being optimized?

a: The most probable sequence of POS tags given the observed words Explanation

Explanation

b: The fastest computational path Explanation

Explanation

c: The shortest tag sequence Explanation

Explanation

d: The random assignment of tags Explanation

Explanation

Which of the following factors would LEAST likely improve POS tagging performance?

a: Adding morphological features (prefixes, suffixes) Explanation

Explanation

b: Including capitalization information Explanation

Explanation

c: Randomly shuffling the word order in training sentences Explanation

Explanation

d: Using word embeddings as additional features Explanation

Explanation

What is the 'label bias problem' that affects some sequence labeling models like HMM?

a: The model prefers states with fewer outgoing transitions Explanation

Explanation

b: The model cannot handle unknown words Explanation

Explanation

c: The model is too slow for real-time applications Explanation

Explanation

d: The model requires too much training data Explanation

Explanation