Virtual Labs

Given a corpus with the sequence 'EOS noun verb EOS', what is the transition probability P(verb|noun)?

a: 0.5 Explanation

Explanation

b: 1.0 Explanation

Explanation

c: 0.33 Explanation

Explanation

d: Cannot be determined from this information alone Explanation

Explanation

If the word 'run' appears 3 times as a verb and 2 times as a noun in a corpus, what is the emission probability P(run|verb)?

a: 0.6 Explanation

Explanation

b: 0.4 Explanation

Explanation

c: Cannot be determined without knowing total verb count Explanation

Explanation

d: 0.5 Explanation

Explanation

What is the significance of the EOS (End of Sentence) marker in HMM-based POS tagging?

a: It increases the vocabulary size Explanation

Explanation

b: It helps model sentence boundaries and initial tag probabilities Explanation

Explanation

c: It makes the algorithm faster Explanation

Explanation

d: It is only used for punctuation Explanation

Explanation

In the Viterbi algorithm, what does the backtracking step accomplish?

a: It calculates the probabilities for each word Explanation

Explanation

b: It finds the sequence of states that led to the highest probability Explanation

Explanation

c: It initializes the algorithm Explanation

Explanation

d: It handles unknown words Explanation

Explanation

Why might the word 'can' be challenging for an HMM-based POS tagger?

a: It's too short Explanation

Explanation

b: It can be both a modal verb and a noun depending on context Explanation

Explanation

c: It appears too frequently Explanation

Explanation

d: It's not in the training data Explanation

Explanation

What happens during the initialization step of the Viterbi algorithm?

a: The most likely final state is determined Explanation

Explanation

b: Probabilities for the first word are calculated using initial state probabilities and emission probabilities Explanation

Explanation

c: The transition matrix is normalized Explanation

Explanation

d: The training data is processed Explanation

Explanation

In a more complex scenario, if we have the sequence 'The/DT dog/NN runs/VBZ fast/RB', what would be the transition probability P(VBZ|NN)?

a: 1.0, since NN is always followed by VBZ in this example Explanation

Explanation

b: 0.5, since there's one transition out of two possible Explanation

Explanation

c: Cannot be determined without more training data Explanation

Explanation

d: 0.25, since there are four words Explanation

Explanation

What is the computational complexity of the Viterbi algorithm for a sentence of length T with N possible POS tags?

a: O(T × N) Explanation

Explanation

b: O(T × N²) Explanation

Explanation

c: O(T²× N) Explanation

Explanation

d: O(N^T) Explanation

Explanation

How does the HMM handle data sparsity problems in POS tagging?

a: By using only the most frequent words Explanation

Explanation

b: By implementing smoothing techniques and handling unknown words Explanation

Explanation

c: By ignoring rare words completely Explanation

Explanation

d: By using only bigram models Explanation

Explanation

POS Tagging - Hidden Markov Model