Hidden Markov Models for Tagging

Hidden Markov Models for Tagging

From Markov chains to Viterbi

Training and decoding a tiny tagger

Which grouping of parts of speech into open and closed classes is right?

A BIO tagging scheme covers $4$ entity types. How many distinct tags does the per-token labeler output? Answer with the number.

In an HMM tagger, what do the A and B tables contain, and how are they estimated?

At the previous column two states have Viterbi values $0.1$ and $0.6$. The transitions into the current state $j$ are $0.3$ and $0.4$, and the emission probability of the current observation in state $j$ is $0.25$. Compute $v_t(j)$, the new cell's Viterbi value, rounding to four decimals.

A tagger has $N = 10$ tags and a sentence of $T = 12$ words. How many cells does the Viterbi lattice fill, and how many complete taggings would exhaustive search examine? Give the number of lattice cells (a single number).