Heinz and Lai (2013): Vowel harmony and subsequentiality #
Heinz and Lai place iterative vowel harmony in the subsequential hierarchy. Over the alphabet of
§3, harmonizing vowels + and −, an inert class C of consonants and transparent vowels, and
opaque vowels ⊞ and ⊟, progressive harmony is computed left to right by the sequential
transducer of Figure 2, and regressive harmony by the same transducer run right to left. Sour
grapes harmony (8) is not left-subsequential (Theorem 5), and dominant–recessive harmony (9) is
subsequential in neither direction (Theorem 6), but it is weakly deterministic in the sense of
their Definition 1 (Theorem 7): progressive spreading of + alone, Figure 4, run left to right
and then right to left.
Main definitions #
Seg: the alphabet of §3ph: progressive harmony, the transducer of Figure 2IsSourGrapes,IsDominantRecessive: the conditions (8) and (9)php: progressive spreading of+alone, the transducer of Figure 4
Main results #
not_isLeftSubsequential_of_isSourGrapes: Theorem 5exists_isRightSubsequential_isSourGrapes,not_isRightSubsequential_of_isSourGrapes: the right half of Theorem 5 fails for (8) alone and holds once trigger-less words are left alonenot_isSubsequential_of_isDominantRecessive: Theorem 6isDominantRecessive_php,isWeaklyDeterministic_php: Theorem 7
Implementation notes #
- The proof of Theorem 5 goes through the bounded delay of left-subsequential functions rather
than the paper's sets of tails. The paper omits the right half as similar; it needs more than
(8), since a right-to-left pass that raises every
−unless the word ends in⊟meets (8). - Theorem 7 is stated, as in the paper, for words over
+and−. - Not formalized: Theorem 4 (majority rules is not regular), the stem-control results (Theorems 8 and 9), and the conjecture that sour grapes is not weakly deterministic.
References #
Equations
- HeinzLai2013.instDecidableEqSeg x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
- HeinzLai2013.instReprSeg.repr HeinzLai2013.Seg.plus prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "HeinzLai2013.Seg.plus")).group prec✝
- HeinzLai2013.instReprSeg.repr HeinzLai2013.Seg.minus prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "HeinzLai2013.Seg.minus")).group prec✝
- HeinzLai2013.instReprSeg.repr HeinzLai2013.Seg.inert prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "HeinzLai2013.Seg.inert")).group prec✝
- HeinzLai2013.instReprSeg.repr HeinzLai2013.Seg.opaquePlus prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "HeinzLai2013.Seg.opaquePlus")).group prec✝
Instances For
Equations
- HeinzLai2013.instReprSeg = { reprPrec := HeinzLai2013.instReprSeg.repr }
Progressive harmony #
Progressive harmony, Figure 2. The state is the value of the current harmonic domain, set by the first harmonizing vowel and reset by each opaque vowel; a harmonizing vowel takes that value. Run right to left, the same machine computes regressive harmony.
Equations
- One or more equations did not get rendered due to their size.
Instances For
ofLetter c is the class of the Turkish letter c. The vowels i and e are [−back],
o and u are [+back], and everything else is inert.
Equations
Instances For
ofAnalysis a is the underlying and the surface string of the word a analyses, read off
its underlying form without the slashes and the morpheme boundary, and off its segments.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Progressive harmony takes the underlying form of each Turkish genitive of Table 2 to its surface form.
Sour grapes #
f meets the sour-grapes condition (8). A trigger followed by −s spreads to all of them, and
does not spread at all when an opaque ⊟ ends the word.
Equations
- One or more equations did not get rendered due to their size.
Instances For
No left-subsequential function meets (8) (Theorem 5). A left scan reading + −ⁿ⁺¹ must hold
back all but a bounded part of its output until it learns whether ⊟ follows.
The right-to-left pass that raises every − unless the word ends in ⊟. Its state records,
after the first symbol read, whether that symbol was ⊟.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The right half of Theorem 5 fails for (8) alone, since a right-to-left pass meets (8).
The right half of Theorem 5 holds once a word with no trigger is left alone. No
length-preserving right-subsequential function meets (8) and leaves every −ⁿ unchanged.
Dominant–recessive harmony #
f meets the dominant–recessive condition (9). On words over + and −, every vowel surfaces
+ if some vowel is +, and − otherwise.
Equations
- One or more equations did not get rendered due to their size.
Instances For
No length-preserving subsequential function, in either direction, meets (9) (Theorem 6). A
medial − among −s turns + under a + placed far to the left or far to the right.
Progressive spreading of + alone (Figure 4) turns every − after a + into +.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Progressive spreading of +, run left to right and then right to left, meets (9)
(Theorem 7).
Dominant–recessive harmony is weakly deterministic (Theorem 7).