S island of guam is maintaining us a high state of

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: glish sentence ˆ E = argmax E P( E | F ) •  The Viterbi approxima1on to the best English sentence: ( A, E ) = argmax ( A, E ) P( A, E | F ) •  Search through the space of all English sentences 61 Dan Jurafsky Phrase ­based Decoding Slide adapted from Philipp Koehn Maria Maria no dio una bofetada a laa bruja verde una bofetada l Mary did not slap the green witch •  Reordering finished to right TranslaBon transla1on Many transla1on lef Build t many One to o one •  Select foreign word to be translated •  Find English phrase transla1on •  Add English phrase to end of par1al transla1on •  Mark foreign words as translated 62 Dan Jurafsky Decoding: The la]ce of possible English transla$ons for phrases • Maria a no dió una bofetada a la bruja verde Mary not give a slap to the witch green did not no a slap to green witch to the slap did not give to the slap 63 the witch Dan Jurafsky Decoding •  Stack decoding •  Maintaining a stack of hypotheses. •  Actually a priority queue •  Actually a set of different priority queues •  Itera1vely pop off the best ­scoring hypothesis, expand it, put back on stack. •  The score for each hypothesis: •  Score so far COST (hyp(S ( E, F )) = " ! ( fi , ei )d (starti # endi#1 # 1)P( E ) •  Es1mate of future costs 64 i! S Dan Jurafsky Decoding by hypothesis expansion Maria no dio una bofetada… •  Afer expanding NULL •  Afer expanding No E: No slap F: *N*UB*** COST: 803 •  Afer expanding Mary 65 Dan Jurafsky Efficiency •  The space of possible transla1ons is huge! •  Even if we have the right n words, there are n! permuta1ons •  We need to find the best scoring permuta1on •  Finding the argmax with an n ­gram language model is NP ­ complet...
View Full Document

Ask a homework question - tutors are online