What is Phonology?
Phonology Pronunciation Spelling
cat + -s “kats” cats
dog + -s “dawgz” dogs
rose + -s “roziz” roses
[These slides are missing most kiss + -s “kisiz” why? kisses
examples and discussion from
class …] How do you pronounce a phonology doesn’t care
about the spelling
sequence of morphemes? (that’s just applied
Especially, how & why do you morphology)
fix up the pronunciation at the
seams between morphemes?
600.465 - Intro to NLP - J. Eisner 1 600.465 - Intro to NLP - J. Eisner 2
What is Phonology? What is Phonology?
Pronunciation Spelling “Trisyllabic Shortening” in English
nap + -t “næpt” napped divine divinity serene serenity
futile futility supreme supremity
nab + -t “næbd” nabbed senile senility obscene obscenity
satire satirical obese *obesity
nod + -t “nadid” nodded decide decision
wild wilderness
knot + -t “natid” knotted
Actually, these are pronounced identically: (and similarly for other vowels)
naId
thanks to the English “flapping” rule
(similarly: ladder/latter, bedding/betting)
600.465 - Intro to NLP - J. Eisner 3 600.465 - Intro to NLP - J. Eisner 4
Successive Fixups
What is Phonology? for Phonology
A function twixt head and lip Chomsky & Halle (1968)
Morphology underlying Phonological surface Articulation
Stepwise refinement of a single form
(head) phonemes mapping phones (mouth) How to handle “resignation” example?
resign ree-ZIYN
resign + -ation reh-zihg-NAY-shun
input (I)
What class of functions is allowed? output (O)
Differs from one language to next
That is, O = f(I) = g3(g2(g1(I)))
Often complicated, but not arbitrary Function composition (e.g., transducer composition)
Comp Sci: How to compute, invert, learn?
600.465 - Intro to NLP - J. Eisner 5 600.465 - Intro to NLP - J. Eisner 6
example courtesy of K. Crosswhite
Optimality Theory
How to Give Orders for Phonology
Directions version: successive fixup (derivation) Prince & Smolensky (1993)
Break two eggs into a medium mixing bowl.
Alternative to successive fixups
Remove this tab first.
On the last day of each month, come to this office Successive winnowing of candidate set
and pay your rent.
Rules version: successive winnowing
No running in the house is allowed. (optimization)
All dogs must be on a leash.
Rent must be paid by the first day of each month.
In rules version, describe what a good solution would input
look like, plus a search procedure for finding the best output
solution). Where else have we seen this?
600.465 - Intro to NLP - J. Eisner 7 600.465 - Intro to NLP - J. Eisner 8
Optimality Theory
Optimality Theory “Tableau”
for Phonology
= candidate violates constraint twice (weight 2)
input (I)
constraint would prefer A, but only output (O)
allowed to break tie among B,D,E
600.465 - Intro to NLP - J. Eisner 9 600.465 - Intro to NLP - J. Eisner 10
When do we prune back to
best paths? Why does order matter?
Optimality Theory: At each intermediate stage Optimality Theory: Each machine (FSA) can choose only
among outputs that previous machines liked best
Noisy channel: After adding up all weights
Noisy channel: Each machine (FST) alters the output
produced by previous machines
output (O) output (O)
600.465 - Intro to NLP - J. Eisner 11 600.465 - Intro to NLP - J. Eisner 12
Final Remark on OT Optimality Theory “Tableau”
Repeated best-paths only works for a single input
Better to build full FST for I O (invertible)
Can do this e.g. if every constraint is binary:
Assigns each candidate either 1 star (“bad”) or 0 stars
(“good”)
input (I)
output (O) all surviving candidates violate constraint 3,
so we can’t eliminate any
600.465 - Intro to NLP - J. Eisner 13 600.465 - Intro to NLP - J. Eisner 14
