The Implementation of the Rules

Next: Runtime Behaviour Up: The rules Previous: Definition of the phoneme

The Implementation of the Rules

The implementation of the individual rules follows.

rule01.fst Stem-initial devoicing
2 states, 10 arcs, Circular.
File length: 278 bytes
This rule was implemented as stated by Kirchhoff.
```
regex [ "(+voice,+cont,apical)" -> "(-voice,+cont,apical)" 
						|| StemInitial_ ];
```

rule02.fst Reduction of affricates

6 states, 44 arcs, Circular.
File length: 391 bytes

This rule was implemented just as stated by Kirchhoff. It consists of two subrules:

r2_1pf.fst For the reduction of /pf/

regex [ "(-voice,affr,labial)" -> "(-voice,+cont,labial)" 
						|| StemInitial_ ];
2 states, 10 arcs, Circular.
File length:      276 bytes

r2_2ts.fst For the reduction of /ts/

regex [ "(-voice,affr,apical)" -> "(-voice,+cont,apical)" 
						|| StemInitial_ ];
2 states, 10 arcs, Circular.
File length:      276 bytes

rule03.fst Glottal Stop Dropping
2 states, 10 arcs, Circular.
File length: 198 bytes
This rule was implemented as stated by Kirchhoff.
```
regex [ "?" -> 0 // Onset_ ];
```

rule04.fst Shortening

255 states, 5351 arcs, Circular.
File length: 16111 bytes

This rule was implemented just as stated by Kirchhoff. It consists of eight subrules:

r4_1i:.fst For the shortening of /i:/

regex [ "(+voice,vowel,[long,high,front,-round])" 
		-> "(+voice,vowel,[high,front,-round])" ];
1 state, 3 arcs, Circular.
File length:      317 bytes

r4_2y:.fst For the shortening of /y:/

regex [ "(+voice,vowel,[long,high,front,+round])" 
-> "(+voice,vowel,[high,front,+round])" ];
1 state, 3 arcs, Circular.
File length:      317 bytes

r4_3e:.fst For the shortening of /e:/

regex [ "(+voice,vowel,[long,mid,front,-round])" 
-> "(+voice,vowel,[mid,front,-round])" ];
1 state, 3 arcs, Circular.
File length:      313 bytes

r4_42:.fst For the shortening of /2:/

regex [ "(+voice,vowel,[long,mid,front,+round])" 
-> "(+voice,vowel,[mid,front,+round])" ];
1 state, 3 arcs, Circular.
File length:      313 bytes

r4_5E:.fst For the shortening of /E:/

regex [ "(+voice,vowel,[long,low,front,-round])" 
-> "(+voice,vowel,[low,front,-round])" ];
1 state, 3 arcs, Circular.
File length:      313 bytes

r4_6u:.fst For the shortening of /u:/

regex [ "(+voice,vowel,[long,high,back,+round])" 
-> "(+voice,vowel,[high,back,+round])" ];
1 state, 3 arcs, Circular.
File length:      313 bytes

r4_7o:.fst For the shortening of /o:/

regex [ "(+voice,vowel,[long,mid,back,+round])" 
-> "(+voice,vowel,[mid,back,+round])" ];
1 state, 3 arcs, Circular.
File length:      309 bytes

r4_8a:.fst For the shortening of /a:/

regex [ "(+voice,vowel,[long,low,back,-round])" 
-> "(+voice,vowel,[low,back,-round])" ];
1 state, 3 arcs, Circular.
File length:      309 bytes

rule04.fst was then produced through union building of all eight subrules. It was implemented one-plus (1+rule04.fst):

xfst[0]: load rule04.fst
xfst[1]: one-plus
1 state, 25 arcs, Circular.
File length:      1379 bytes

regel05.fst Laxing
1 state, 3 arcs, Circular.
File length: 311 bytes
This rule was implemented only for the substituion /e:/ /E/, the other cases stated by Kirchhoff were not considered.
```
regex [ "(+voice,vowel,[long,mid,front,-round])" 
		-> "(+voice,vowel,[low,front,-round])" ];
```

rule06.fst E: - e: Alternation
6 states, 124 arcs, Circular.
File length: 2139 bytes
This rule was implemented through the union building of two subrules, r6_1E:.fst and r6_2E.fst. The first one corresponds to the definition of the rule given by Kirchhoff in her rule 6. The second subrule is a completion of it, as stated in the explanatory text of her rule 11. The second subrule does not apply within a diphthong.
- r6_1E:.fst For the /E://e:/ Alternation
```
 regex [ "(+voice,vowel,[long,low,front,-round])" \\
-> "(+voice,vowel,[long,mid,front,-round])" ];
1 state, 3 arcs, Circular.
File length:      323 bytes
```
- r6_2E.fst For the /E//e/ Alternation (stated in rule 11)
```
regex [ "(+voice,vowel,[low,front,-round])"\\
 -> "(+voice,vowel,[mid,front,-round])"\\
|| _Vowel ];
3 states, 50 arcs, Circular.
```

rule07.fst I-Rounding

4 states, 14 arcs, Circular.
File length: 417 bytes

This rule was specified as stated in Kirchhoff's report [Kir95]:

regex [ "(+voice,vowel,[high,front,central,-round])" 
	-> "(+voice,vowel,[high,front,central,+round])" 
			|| ?_"(+voice,vowel,[low,central])" ];

rule08.fst Schwa-reduction
4 states, 112 arcs, Circular.
File length: 2176 bytes
This rule was specified as stated in Kirchhoff's report [Kir95]:
```
regex [ Vowel - Central -> "(+voice,vowel,[mid,central])" 
						|| \Stress_\Vowel ];
```
This rule was implemented one-plus (1+rule08.fst), so that it can change the input string at any potentially possible position:
```
xfst[0]: load rule08.fst
xfst[1]: one-plus
3 states, 72 arcs, Circular.
File length:      2112 bytes
```

rule09.fst Diphthong reduction: 6-dropping
4 states, 74 arcs, Circular.
File length: 2061 bytes
This rule was specified as stated in Kirchhoff's report [Kir95]:
```
regex [ "(+voice,vowel,[low,central])" -> 0 
					|| Vowel_? ];
```
This rule was implemented one-plus (1+rule09.fst), so that it can change the input string at any potentially possible position:
```
xfst[0]: load rule09.fst
xfst[1]: one-plus
3 states, 74 arcs, Circular.
File length:      2061 bytes
```

rule10.fst Diphthong reduction: reduction to schwa
4 states, 121 arcs, Circular.
File length: 2213 bytes
This rule was implemented through the composition of two subrules: the first one, r10_16.fst, for the substituion /6/ /@/, and the second one, r10_2del.fst, for the deletion of the vowel.
- r10_16.fst for the substitution /6/ /@/
```
regex [ "(+voice,vowel,[low,central])" -> "(+voice,vowel,[mid,central])" \\
|| \PrimStress Vowel_ ];
3 states, 75 arcs, Circular.
File length:      2069 bytes
```
- r10_2del.fst for the deletion of the vowel
```
regex [ Vowel -> 0 || \PrimStress_"(+voice,vowel,[mid,central])" ];
4 states, 121 arcs, Circular.
File length:      2250 bytes
```
This rule was implemented one-plus (1+rule10.fst), so that it can process the input string at any potentially possible position:
```
xfst[0]: load rule10.fst
xfst[1]: one-plus
4 states, 123 arcs, Circular.
File length:      2213 bytes
```

rule11.fst Diphthong reduction: reduction to a-schwa
4 states, 29 arcs, Circular.
File length: 580 bytes
For the implementation of this rule, a special class was created, ReductionASchwa, which containing exactly those phonemes that are affected by the rule. They are listed in the cited report.
```
regex [ ReductionASchwa -> 0 // \PrimStress_ASchwa ];
```

rule12.fst Nasal assimilation
4 states, 24 arcs, Circular.
File length: 698 bytes
This rule was implemented through union building over two subrules: r12_1nm.fst, for the labial nasals and r12_2nN.fst, for the velar nasals.
- r12_1nm.fst for the labial nasals
```
regex [ "(+voice,nasal,apical)" 
		-> "(+voice,nasal,labial)" // _Labial.#. ];
5 states, 32 arcs, Circular.
File length:      535 bytes
```
- r12_2nN.fst for the velar nasals
```
regex [ "(+voice,nasal,apical)" 
		-> "(+voice,nasal,velar)" || _Back.#. ];
5 states, 28 arcs, Circular.
File length:      474 bytes
```

rule13.fst Plosive epenthesis
6 states, 61 arcs, Circular.
File length: 780 bytes
This rule was implemented through union building of three subrules: the first one, r13_1mpf.fst, for the labial plosives, the second one, r13_2nts.fst, for the apical plosives and the third one, r13_3Nks.fst, for the velar plosives .
- r13_1mpf.fst for the labial plosives
```
regex [ 0 -> "(-voice,-cont,labial)" // m_MinusVoiPlusCont.#. ];
4 states, 25 arcs, Circular.
File length:      571 bytes
```
- r13_2nts.fst for the apical plosives
```
regex [ 0 -> "(-voice,-cont,apical)" // n_MinusVoiPlusCont.#. ];
4 states, 25 arcs, Circular.
File length:      569 bytes
```
- r13_3Nks.fst for the velar plosives
```
regex [ 0 -> "(-voice,-cont,velar)" // N_MinusVoiPlusCont.#. ];
4 states, 25 arcs, Circular.
File length:      565 bytes
```

rule14.fst Plosive Deletion
5 states, 90 arcs, Circular.
File length: 1282 bytes
Rule 14 was implemented through union building of two subrules. The first one, r14_1all.fst, the most general one, stands for the deletion of all plosives between a nasal and a continuant consonant. The second one, r14_2nts.fst, is for the specific case /nts//ns/. The extra specification of this particular case was necessary, as /ts/ is declared as one phoneme and is therefore recognized by the XFST system as an atom. However, the union of both these rules is equivalent to the declaration of the rule in the report [Kir95]
- r14_1all.fst the general rule
```
regex [ ContMinus -> 0 // Nasal_ContPlus.#. ];
6 states, 101 arcs, Circular.
File length:      1292 bytes
```
- r14_2nts.fst for the specific case /nts//ns/
```
regex [ "(-voice,affr,apical)" -> "(-voice,+cont,apical)" 
				|| "(+voice,nasal,apical)"_.#. ];
4 states, 13 arcs, Circular.
File length:      316 bytes
```

Rule 15 was not implemented in the final version of the automaton.

rule16.fst R-Alternation

6 states, 104 arcs, Circular.
File length: 2176 bytes

regex [ "(+voice,vowel,[low,central])" 
		-> "(+voice,+cont,uvular)" 
				// Vowel_\Vowel.#. ];

rule17.fst Velar fricative weakening
7 states, 137 arcs, Circular.
File length: 2302 bytes
```
regex [ "(-voice,+cont,velar)" 
	-> "(-voice,+cont,glottal)" 
			// ?_\Vowel.#. ];
```

regel18.fst L-vocalisation

4 states, 25 arcs, Circular.
File length: 669 bytes

regex [ "(+voice,lateral,apical)" 
	-> "(+voice,vowel,[high,front,central,-round])" 
				|| PlusLowPlusFront_FrontPlus ];

rule19.fst Schwa-deletion
6 states, 130 arcs, Circular.
File length: 3251 bytes
This rule was implemented stem-final, as stated in Kirchhoff's report [Kir95].
```
regex [ "(+voice,vowel,[mid,central])" -> 0 
			|| Consonant_Nasal.#. | Lateral.#. ];
```

rule20.fst Schwa-deletion and progressive place assimilation
12 states, 333 arcs, Circular.
File length: 3251 bytes
Rule 20 was implemented through composition of two subrules: r20_1Schwa.fst for the Schwa-deletion and r20_2Assim.fst for the progressive place assimilation.
- r20_1Schwa.fst Schwa-deletion
```
regex [ "(+voice,vowel,[mid,central])" -> 0 
				|| Consonant_Nasal.#. ];
6 states, 125 arcs, Circular.
File length:      2341 bytes
```
- r20_2Assim.fst Progressive place assimilation
  4 states, 41 arcs, Circular.
  File length: 732 bytes
  
  r20_2Assim.fst was produced through union building over two subrules:
  - r20_2aLabial.fst for the labials
```
regex [ "(+voice,nasal,apical)" 
		-> "(+voice,nasal,labial)" || Labial_.#. ];
4 states, 25 arcs, Circular.
File length:      520 bytes
```
  - r20_2bVelar.fst for the velars
```
regex [ "(+voice,nasal,apical)" 
		-> "(+voice,nasal,velar)" || Velar_.#. ];
4 states, 22 arcs, Circular.
File length:      463 bytes
```

rule21.fst Schwa-deletion and regressive place assimilation 11 states, 48 arcs, Circular.
File length: 401 bytes
Rule 21 was implemented through composition of two subrules: r21_1Schwa.fst for the Schwa-deletion and r21_2Assim.fst for the regressive place assimilation.
- r21_1Schwa.fst Schwa-deletion
```
 regex [ "(+voice,vowel,[mid,central])" -> 0 || Nasal_Nasal.#. ];
6 states, 25 arcs, Circular.
File length:      397 bytes
```
- r21_2Assim.fst Schwa-deletion
```
regex [ "(+voice,nasal,apical)" -> "(+voice,nasal,labial)" 
				|| _"(+voice,nasal,labial)".#. ];
5 states, 13 arcs, Circular.
File length:      272 bytes
```
rule21.fst was then produced through composition over the two subrules.

rule22.fst Schwa-deletion and consonant assimilation
12 states, 87 arcs, Circular.
File length: 949 bytes
Rule 22 was implemented with the union building of two rules, rule22a.fst and rule22b.fst, as stated in Kirchhoff's report [Kir95]. The first rule is for the deletion of one [+voi,-cont] consonant, the second one for the deletion of the /n/. The rule was implemented stem-final.
- rule22a.fst Schwa-deletion and consonant assimilation
  15 states, 173 arcs, Circular.
  File length: 1125 bytes
  Rule 22(a) was implemented through composition of three subrules: r22a_1Schwa.fst, r22a_2Assim.fst and r22a_3Consonant.fst.
  - r22a_1Schwa.fst for the Schwa-deletion
```
regex [ "(+voice,vowel,[mid,central])" 
-> 0 || PlusVoiMinusCont_Nasal.#. ];
6 states, 36 arcs, Circular.
File length:      552 bytes
```
  - r22a_2Assim.fst for the assimilation of the consonant
    4 states, 41 arcs, Circular.
    File length: 732 bytes
    This subrule was implemented through union bulding of two subrules: r22a_2Labial.fst and r22a_2Velar.fst.
    - r22a_2Labial.fst for the assimilation of /n/ to a labial consonant
```
regex [ "(+voice,nasal,apical)" 
-> "(+voice,nasal,labial)" || Labial_.#. ];
4 states, 25 arcs, Circular.
File length:      520 bytes
```
    - r22a_2Velar.fst for the assimilation of /n/ to a velar consonant
```
regex  [ "(+voice,nasal,apical)"
 -> "(+voice,nasal,velar)" || Velar_.#. ];
4 states, 22 arcs, Circular.
File length:      520 bytes
```
    Through union building r22_2Assim was created.
  - r22a_3Consonant.fst for the consonant deletion
```
regex [ PlusVoiMinusCont -> 0 || _Nasal.#. ];
5 states, 33 arcs, Circular.
File length:      492 bytes
```
Through composition of the three subrules rule22a.fst was produced.
- rule22b.fst for the /n/-deletion
  7 states, 23 arcs, Circular.
  File length: 312 bytes
  
  rule 22(b) was produced through the compostion of two subrules: r22b_1Schwa.fst for the Schwa-deletion and r22b_2n.fst for the deletion of the /n/ before a /m/.
  - r22b_1Schwa.fst for the Schwa-deletion
```
regex [ "(+voice,vowel,[mid,central])" -> 0  
|| "(+voice,nasal,apical)"_"(+voice,nasal,labial)".#. ];
6 states, 18 arcs, Circular.
File length:      340 bytes
```
  - r22b_2n.fst for the /n/-deletion
```
regex [ "(+voice,nasal,apical)" -> 0 
|| _"(+voice,nasal,labial)".#. ];
5 states, 13 arcs, Circular.
File length:      272 bytes
```
  Through composition of the two subrules rule22b.fst was produced.
rule22.fst was then produced through union building of rule22a.fst and rule22b.fst.

rule23.fst Schwa-deletion and consonant assimilation
16 states, 147 arcs, Circular.
File length: 919 bytes
Rule 23 was implemented with the composition of three subrules: the first one for the schwa-deletion, the second one for the assimilation of the nasal and the third one for the nasalisation of the consonant.
- r23_1Schwa.fst for the schwa-deletion
```
regex [ "(+voice,vowel,[mid,central])" -> 0 
				|| ContMinus_Nasal.#. ];
6 states, 48 arcs, Circular.
File length:      709 bytes
```

r23_2Assim.fst for the assimilation of the nasal

5 states, 46 arcs, Circular.
File length: 618 bytes

This subrule was built through union building over three subrules:

r23_2N.fst for the schwa-deletion

regex [ Nasal -> "(+voice,nasal,velar)" 
|| MinusContPlusVel_.#. ];
4 states, 20 arcs, Circular.
File length:      418 bytes

r23_2n.fst for the schwa-deletion

regex [ Nasal -> "(+voice,nasal,apical)" 
|| MinusContPlusApi_.#. ];
4 states, 20 arcs, Circular.
File length:      422 bytes

r23_2m.fst for the schwa-deletion

regex [ Nasal -> "(+voice,nasal,labial)" 
|| MinusContPlusLab_.#. ];
4 states, 20 arcs, Circular.
File length:      422 bytes

r23_3Nasal.fst for the assimilation of the nasal
5 states, 19 arcs, Circular. File length: 564 bytes
This subrule was built through union building over three subrules:
- r23_3aLabial.fst for the schwa-deletion
```
 regex [ MinusContPlusLab ->  "(+voice,nasal,labial)" 
|| _"(+voice,nasal,labial)".#. ];
5 states, 19 arcs, Circular.
File length:      331 bytes
```
- r23_3bApical.fst for the schwa-deletion
```
regex [ MinusContPlusApi -> "(+voice,nasal,apical)" 
|| _"(+voice,nasal,apical)".#. ];
5 states, 19 arcs, Circular.
File length:      331 bytes
```
- r23_3cVelar.fst for the schwa-deletion
```
regex [ MinusContPlusVel -> "(+voice,nasal,velar)" 
|| _"(+voice,nasal,velar)".#. ];                 
5 states, 19 arcs, Circular.
File length:      325 bytes
```
The subrules r23_2Assim.fst and r23_3Nasal.fst were each built through union building of the corresponding subrules listed above. rule23.fst was then implemented with composition of the three subrules, r23_1Schwa.fst, r23_2Assim.fst and r23_3Nasal.fst in this order.

rule24.fst Voicing assimilation

709 states, 30634 arcs, Circular.
File length: 107534 bytes

Rule 24 was implemented through union building of seven subrules, which follow:

r24_1pb.fst for the substitution /p/ /b/

regex [ "(-voice,-cont,labial)" 
-> "(+voice,-cont,labial)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2761 bytes

r24_2td.fst for the substitution /t/ /d/

regex ["(-voice,-cont,apical)" 
-> "(+voice,-cont,apical)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2761 bytes

r24_3kg.fst for the substitution /k/ /g/

regex [ "(-voice,-cont,velar)" 
-> "(+voice,-cont,velar)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2759 bytes

r24_4fv.fst for the substitution /f/ /v/

regex [ "(-voice,+cont,labial)" 
-> "(+voice,+cont,labial)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2761 bytes

r24_5sz.fst for the substitution /s/ /z/

regex [ "(-voice,+cont,apical)" 
-> "(+voice,+cont,apical)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2761 bytes

r24_6SZ.fst for the substitution /S/ /Z/

regex [ "(-voice,+cont,palato)" 
-> "(+voice,+cont,palato)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2761 bytes

r24_7Cj.fst for the substitution /C/ /j/

regex ["(-voice,+cont,palatal)" 
-> "(+voice,+cont,palatal)" || PlusVoi_PlusVoi ];
4 states, 112 arcs, Circular.
File length:      2763 bytes

Rule24.fst was then built through union buiding over all seven subrules.

Rule 25 (Laxing of Constriction: labials) and rule 26 (Laxing of Constriction: velars) were not implemented.

rule27.fst Cross-syllable plosive assimilation/deletion
4 states, 69 arcs, Circular.
File length: 1224 bytes
From rule 27 only the subrule (c) (as described in [Kir95]) was implemented (Deletion before fricatives). This subrule is described below:
- r27_c.fst
```
regex [ ContMinus -> 0 || Nasal_ContPlus ];
```

Rule 28 (Cluster simplification) was not implemented.

rule29.fst Schwa-Lowering

4 states, 85 arcs, Circular.
File length: 2153 bytes

regex [ "(+voice,vowel,[mid,central])" 
-> "(+voice,vowel,[low,central])" 
|| Consonant_"(+voice,+cont,uvular)" ];

rule30.fst Stem-final mixed junction

891 states, 24845 arcs, Circular.
File length: 83721 bytes

This rule was implemented here consisting of three subrules, which were then composed: the first rule is the /r/ deletion, the second rule is the substitution of /@/ into /6/ an the third is the substitution of the long vowels by central ones. This last subrule is itself built through union building of five subrules.

r30_1r.fst for the /r/ deletion

regex [ "(+voice,+cont,uvular)" -> 0 
// LongVowel_Schwa "(+voice,nasal,apical)" .#. ];
8 states, 63 arcs, Circular.
File length:      1060 bytes

r30_2@.fst for the substitution of /@/ by /6/

regex [ "(+voice,vowel,[mid,central])" -> "(+voice,vowel,[low,central])" 
|| LongVowel_"(+voice,nasal,apical)" ];
4 states, 37 arcs, Circular.
File length:      1025 bytes

r30_3ai:.fst for the substitution of /i:/ by /I/

regex [ "(+voice,vowel,[long,high,front,-round])" -> 
"(+voice,vowel,[high,front,central,-round])" 
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

r30_3by:.fst for the substituion of /y:/ by /Y/

regex [ "(+voice,vowel,[long,high,front,+round])" -> 
"(+voice,vowel,[high,front,central,+round])" 
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

r30_3cu:.fst for the substituion of /u:/ by /U/

regex [ "(+voice,vowel,[long,high,back,+round])" -> 
"(+voice,vowel,[high,back,central,+round])" 
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

r30_3do:.fst for the substituion of /o:/ by /O/

regex [ "(+voice,vowel,[long,mid,back,+round])" -> 
"(+voice,vowel,[mid,back,central,+round])" 
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

r30_3eE:.fst for the substitution of /E:/ by /E/

regex [ "(+voice,vowel,[long,low,front,-round])" -> 
"(+voice,vowel,[low,front,-round])"  
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

r30_3f2:.fst for the substitution of /2:/ by /9/

regex [ "(+voice,vowel,[long,mid,front,+round])" 
-> "(+voice,vowel,[mid,front,central,+round])" 
|| Consonant_"(+voice,vowel,[low,central])" "(+voice,nasal,apical)" ];
6 states, 114 arcs, Circular.
File length:      2221 bytes

The rule r30_3central.fst was generated through union building of the rules r30_3(...).fst defined above.

522 states, 14478 arcs, Circular.
File length: 48258 bytes

rule30.fst was then implemented composing the rules r30_1r.fst, r30_2@.fst and r30_3central.fst.

rule31.fst Stem-final cluster simplification
9 states, 43 arcs, Circular.
File length: 360 bytes
In Kirchhoff's report, this rule serves the schwa-deletion and cluster reduction taking place in one pass. It was implemented there like that because the Two-Level formalism does not permit rules to apply sequentially. But working with finite state transducers permits us to compose two defined rules, forcing them to apply sequentially. This is the reason why rule31.fst was implemented here consisting of two subrules: the first one for the cluster reduction and the second one for the schwa-deletion. The order of application of the subrules doesn't actually matter in this case.
- r31_1t.fst for the cluster reduction (/t/ deletion)
```
regex [ "(-voice,-cont,apical)" -> 0 
|| "(-voice,+cont,apical)"_"(+voice,vowel,[mid,central])" ];
6 states, 29 arcs, Circular.
File length:      371 bytes
```
- r31_2@.fst for schwa deletion
```
regex [ "(+voice,vowel,[mid,central])" -> 0 || "(-voice,+cont,apical)"_? ];
5 states, 21 arcs, Circular.
File length:      312 bytes
```
rule31.fst was then implemented composing both subrules defined above.

Rule 32 (Velar reinforcement) was not implemented because of its strong dialectal character.

rule33.fst Velar fricativisation

13 states, 301 arcs, Circular.
\newline File length:      2592 bytes

This rule was implemented stem-final, as stated. It consists of two subrules: the first one for the back vowels, the second one for the front vowels. The subrules were implemented as follows:

r33_1back.fst for the back vowels

regex [ "(-voice,-cont,velar)" -> "(-voice,+cont,velar)" 
|| BackVowel_StemInitial ];
4 states, 40 arcs, Circular.
File length:      936 bytes

r33_2front.fst for the back vowels

regex [ "(-voice,-cont,velar)" -> "(-voice,+cont,palatal)" 
|| FrontVowel_StemInitial ];
4 states, 55 arcs, Circular.
File length:      1380 bytes

rule33.fst was implemented through union building over both subrules.

rule34.fst Final /C/ /k/ alternation

This rule was implemented stem-final, as required by Kirchhoff. In order to avoid overgeneration of forms like

?'E:n.+lIk

(for ähnlich), it was restricted to the suffix +ig/+IC

regex [ "(-voice,+cont,palatal)" -> "(-voice,-cont,velar)" 
|| "+" "(+voice,vowel,[high,front,central,-round])"_StemInitial+ ];
5 states, 25 arcs, Circular.
File length:      383 bytes

Rule 35 (assimilation of voicelessness) was left out when building of the final version of x-vario.

rule36.fst Cross-syllable nasal assimilation
13 states, 155 arcs, Circular.
File length: 967 bytes
This rule was implemented consisting of two subrules: the first one for the labial nasals, the second one for the velar nasals. The apical nasal could be omitted here, as no substitution takes place for it. The subrules were implemented as follows:
- r36_1labial.fst for the labial nasals
```
regex [ "(+voice,nasal,apical)" -> "(+voice,nasal,labial)" 
					|| _StemBoundary Labial ];
5 states, 37 arcs, Circular.
File length:      552 bytes
```
- r36_2velar.fst for the velar nasals
```
regex [ "(+voice,nasal,apical)" -> "(+voice,nasal,velar)" 
					|| _StemBoundary Velar ];
5 states, 34 arcs, Circular.
File length:      495 bytes
```
rule36.fst was then implemented through union building of the two subrules.

Rules 37 (Geminate simplification) was not implemented, because it applies over words boundaries and is therefore irrelevant for the purposes of processing lexical representations. Rule 38 (place assimilation of /t/) was not implemented because it seems to occur only seldom.

Thorsten Trippel Mon Mar 20 14:35:50 MET 2000