FinnishOTProsody - Fi(n)nish OT

Fi(n)nish OT Prosody
Fi(n)nish OT Prosody
Lauri Karttunen
FSMNLP-2005
September 1, 2005

Overview
Overview
Success of Finite-State Morphology
Lexical transducers
Two ways of describing morphological alterations
Sequential (Chomsky & Halle 1968)
Parallel (Koskenniemi 1983)
Optimality Theory (Prince & Smolensky 1993)
Yet another two-level model
Finite-state?
Lenient composition
Finnish OT Prosody
Basic Facts
Finite-state implementation of Kiparsky’s 2003 analysis
with the FST tool (Beesley & Karttunen 2003)
Conclusion
Final thoughts

Computational morphology
Computational morphology
Analysis
leaves
leaf N Pl leave N Pl leave V Sg3
Generation
hang V Past
hanged hung

Two challenges
Two challenges
Morphotactics
Words are composed of smaller elements that
must be combined in a certain order:
piti-less-ness is English
piti-ness-less is not English
Phonological alternations
The shape of an element may vary depending on
the context
pity is realized as piti in pitilessness
die becomes dy in dying

Morphology is regular (=rational)
Morphology is regular (=rational)
The relation between the surface forms of a language
and the corresponding lexical forms can be described
as a regular relation .
A regular relation consists of ordered pairs of strings.
leaf+N+Pl : leaves hang+V+Past : hung
Any finite collection of such pairs is a regular relation.
Regular relations are closed under operations such as
concatenation , iteration , union , and composition .
Complex regular relations can be derived from simple
relations.

Morphology is finite-state
Morphology is finite-state
A regular relation can be defined using the
metalanguage of regular expressions .
[{talk}|{walk}|{work}] [% +Base:0 | %+SgGen3:s|
%+Progr:{ing}| %+Past:{ed}];
A regular expression can be compiled into a
finite-state transducer that implements
the relation computationally.

work+3rdSg --> works
k : k
t:t
a:a
a:a
w : w
o : o
l:l
r : r
+Progr:i
ε :g
+3rdSg : s
+Past:e
ε :d
ε :n
+Base: ε
Generation
Generation

talked --> talk+Past
k : k
t : t
a:a
a : a
w:w
o:o
l : l
r:r
+Progr:i
ε :g
+3rdSg:s
+Past : e
ε : d
ε :n
+Base: ε
Analysis
Analysis

Lexical transducer
Lexical transducer
veut
vouloir +IndP +SG + P3
Finite-state
transducer
inflected form
citation form inflection codes
v o u l o i r +IndP +SG +P3
v e u t
Bidirectional: generation or analysis
Compact and fast
Comprehensive systems have been
built for over 40 languages:
English, German, Dutch, French,
Italian, Spanish, Portuguese,
Finnish, Russian, Turkish,
Japanese, Korean, Basque, Greek,
Arabic, Hebrew, Bulgarian, …

How lexical transducers are made
How lexical transducers are made
Lexicon
FST
Rule
FSTs
Compiler
f a t
+Adj
r
+Comp
f a t
t e
Lexical Transducer
(a single FST)
composition
Lexicon
Regular Expression
Rules
Regular Expressions
Morphotactics
Alternations

Two-level rules vs. rewrite rules
Two-level rules vs. rewrite rules
co
mpose
intersect
FST
rule 1 rule 2
rule n ...
Surface form
Lexical form
Koskenniemi 1983
Intermediate form
...
Surface form
Lexical form
rule 1
rule n
rule 1
Chomsky&Halle 1968

Rewrite rules
Rewrite rules
Epenthesis
Harmony
Lowering
? u: t y ? A s
? u: t I y ? A s
? u: t u y ? a s
? o: t u y ? a s
Yawelmani
Vowel Harmony
Kisseberth 1969