Lecture 7 HMMs – the 3 Problems Forward Algorithm CSCE 771 Natural Language Processing Lecture 7 HMMs – the 3 Problems Forward Algorithm Topics Overview Readings: Chapter 6 February 6, 2013
Overview Last Time Today Tagging Markov Chains Hidden Markov Models NLTK book – chapter 5 tagging Today Viterbi dynamic programming calculation Noam Chomsky on You Tube Revisited smoothing Dealing with zeroes Laplace Good-Turing
Katz Backoff
Back to Tagging Brown Tagset - In 1967, Kucera and Francis published their classic work Computational Analysis of Present-Day American English – tags added later ~1979 500 texts each roughly 2000 words Zipf’s Law – “the frequency of the n-th most frequent word is roughly proportional to 1/n” Newer larger corpora ~ 100 million words Corpus of Contemporary American English, the British National Corpus or the International Corpus of English http://en.wikipedia.org/wiki/Brown_Corpus
Figure 5.4 pronoun in Celex Counts from COBUILD 16-million word corpus
Figure 5.6 Penn Treebank Tagset
Figure 5.7
Figure 5.7 continued
Figure 5.8
Figure 5.10
5.5.4 Extending HMM to Trigrams Find best tag sequence Bayes rule Markov assumption Extended for Trigrams
Chapter 6 - HMMs formalism revisited
Markov – Output Independence Markov Assumption Output Independence: (Eq 6.7)
Figure 6.2 initial probabilities
Figure 6.3 Example Markov chain Probability of a sequence
Figure 6.4 Probability zero links (Bakis model for temporal problems)
HMMs – The Three Problems
Likelihood Computation – The Forward Algorithm Computing Likelihood: Given an HMM λ = (A, B) and an observation sequence O = o1, o2, … ot, determine the likelihood P(O | λ)
Figure 6.5 B – observational Probabilities for 3 1 3 ice creams
Figure 6.6 transitions for 3 1 3 ice creams
Likelihood computation
Figure 6.7 forward computation
Figure 6.8
Figure 6.9 Forward Algorithm
Figure 6.10
Figure 6.11
Figure 6.12
Figure 6.13
Figure 6.14