Lule Sami NLP Grammar

Finite state and Constraint Grammar based analysers, proofing tools and other resources

View the project on GitHub giellalt/lang-smj

Lule Sámi morphophonological rule set

This file documents the phonology.twolc file

The file contains the rule set for the non-segmental Lule Sámi morphphonological rules

Background

The file is modeled upon the corresponding file for North Sámi, but has been revised and differs from it on several issues. The grammatical sources are Spiik 1989: Lulesamisk grammatik and Nystø and Johnsen 2001: Sámásta 2.

The rule file has the sections Alphabet, Sets, Definition and Rules. The rules are ordered thematically, with 3 main sections: Consonant alternations (except CG), vowel alternations, and consonant gradation.

Declarations and definitions

The Alphabet section

The real Lule Sámi Alphabet

All Lule Saami letters are listed. The Lule Sámi ENG sound is represented as ñ. Lule Sámi letter repertoire is not fully standardised. In the source code we write (and you shall write!) æ; ø; ŋ, but the parser tolerates input written with the the letters ä; ö; ń, ñ (cf. the 4 rules in the file smj/src/orthography/spellrelax.regex).

The 3rd degree mark º is never realized, hence declared as º:0. º:0 = Gradation mark %/ = Literal /, not the TWOLC reserved symbol ‘:’ = Apostrophe

Literal quotes and angles must be escaped (cf morpheme boundaries further down):

h2, g2 etc. are consonants deleted in the Nom. m3, d3 etc. (?) are consonants that undergo certain processes word-finally. This issue should be looked into. Perhaps the two sets can be unified. The reason why there are more distinctions than for sme, is that the cns deletion process is more phonological in sme.

The Dummy symbols

The Dummy symbols are taken from the sme file for convenience, only a small part of them are actually used, they are defined in the Sets section along the way, included there as soon as they are used. The set of actually used Dummy symbols is thus the set declared in “Dummy”. The Dummy symbols trigger morphophonological rules. X is used for nouns and adjectives, Y for verbs and Q for processes common to all The symbols themselves are used in the following way:

OBS: the definitions are not all correct or sufficiently specific

Morpheme boundaries:

The Sets section

These are the sets:

The Definitions section

In this section, the consonants are defined. This includes consonant clusters in the various grades and consonant alternations.

G3 vs G2

The alternation patterns according to Spiik’s alternations series, here named S4, S5, … for “Spiik alternation series 4, 5, etc.” as they are presented in his grammar..

Class Alternation Series
S7 kkn:k0n series 1
S8 fºf:f0f series 2
S9 jgg:j0g series 3
S4 hkk:h0k series 4
S5 xy:zy (no zeros) series 5
S6 xx:yy (no zeros) series 6
S7 xy:zy (no zeros) series 7
S8 —– (no cg) series 8

Definition of gradation symbols:

The Rules section

Overview

The rules section has the following chapters: Consonant alternations in certain pos, vowel lengthening, diphthong simplification, stem vowel alternations, consonant gradation rules

Consonant alternations in certain pos

All rules deal with word-final position.

**Word Final Devoicing of Certain Single Consonants d9 etc. **

**Word Final Devoicing of Certain Single Consonants m9-v ** ! Spilt up because of err/orths ending on v, gierkav> we want err/orth gierkkam

**Err/Orths. **

Word final weakening -tj and -ttj to -sj part 1

Word final weakening -tj and -ttj to -sj part 2

Word Final Deletion of n8 m8 g8 h8

Word Final Neutralization of g8, h8, m8

Deleting Final h9 in Short Essive of Uneven Syllables

Deleting Final l9 in Short Essive of Uneven Syllables

Deleting Final m9 in Short Essive of Uneven Syllables

Deleting Final n9 in Short Essive of Uneven Syllables

Deleting Final r9 in Short Essive of Uneven Syllables

Vowel lengthening

The second syllable vowel a is lengthened to á whenever the stem consonants are in grade 1 and the first syllable vowel is short. Short vowels cannot preceed and follow a single intervocalic consonant.

Compulsatory lengthening in grade I even-syllables

Diphtong simplification

The diphthong simplification handles oa:å and æ:e. Phonologically, these are identical processes, but since the dipthong is written by two letters in the former case and by one letter in the latter, the alternations must be handled separately. This section also handles ie:æ, these are in principle the same as oa:å, but the alternation does not occur in so many contexts.

**oa:å Diphtong Simplification Part I **

oa:å Diphtong Simplification Part II

**æ:e Diphthong Simplification **

**ie:æä Diphthong Simplification Part I **

ie:æä Diphthong Simplification Part II The multichar æä is always the only option

Vowel-change oa:å for verbs part I

Vowel-change oa:å for verbs part II

Stem vowel alternations

This section is divided according to stem vowels: a-, e-, o-, å-stems.

a-stem alternations

For a-stems, there is a:e and a:i. Each alternation is triggered by a combination of phonological content and dummy symbols.

a:e in Present Participle of even-syllable verbs

a:i in Prs Prc of even-syllable verbs

a-stem vowel deletion

e-stem alternations

For e-stems, there is e:i, e:á, e:å, e:u and e:a. Each alternation is triggered by a combination of phonological content and dummy symbols.

e:i in e-stems

The following two rules constitute a <= / => rule pair.

e:á in certain stem types 1

e:á in certain stem types 2

e:å in certain stem types with å as root vowel

e-stem vowel deletion

i-stem alternations

For i-stems, there is i:á. The alternation is triggered by a combination of phonological content and dummy symbols.

i:á in Verb Derivation

o-stem alternations

The duplicates of the three lines of the two following rules are there to resolve the => conflict between the two rules.

o:u in certain stem types 1

o:u in certain stem types 2

u:o in contracted nouns

o-stem vowel deletion

For å-stems there is å:e and å:i and vowel deletion. Each alternation is triggered by a combination of phonological content and dummy symbols.

å:e in Present Participle of even-syllable verbs

å:i in Actor nouns of even-syllable verbs

å-stem vowel deletion

alternations valid for several stem types

Stem vowel deletion in even-syllable verbs, imp 3sg, 3du, 2pl, 3pl

Consonant gradation rules

The consonant gradation rules differ considerably from the corresponding rules for North Sámi. Instead of generalizing oversets of consonants (Cx:Cy <=> …), each rule contains the alternation for one consonant only, and to the right of the <=> arrow is listed all the contexts where the relevant alternation appears. The disadvantage with this method is that the same context must be written several times, if e.g. both p, t and k are deleted in the same contexts, each of these contexts must be written several times, one for each consonant. The advantage is that there are no conflicts during compilation, compilation takes 10 seconds rather than 3 minutes. The earlier North-Sámi-style rule set was ordered according to CG pattern. This pattern is still visible in the new rules, via the reference S1-3 etc. (Spiik’s Series 1, 3-letter pattern, etc) behind each subrule.

This actually opens up for a migration to an xfst rule file instead of the current twolc format, since what xfst really cannot do is generalize over sets (Cx:Cy etc.). This is an issue for future revisions to decide.

The rules are divided in two subsections, deletion rules and change (alternation) rules.

Deletion rules

The b, d, g deletion rules are similar, via the optional ( b ) etc. in front of the “_” symbol, both bm:m and bbm:bm alternations are covered. The contexts differ to a certain extent. For b and d, the III-I special gradation bbm:m is covered by two separate rules, and a special Dummy (X6), not part of the ordinary WeG set.

Note that one of the rules for t:0 refers to #: as part of its context. As soon as clitics are added to the word form, this rule will thus not be triggered. Look into this when the clitics are added.

Consonant gradation b:0 deletes b in S7 and S9 contexts

Consonant gradation d:0 … etc.

Consonant gradation g:0

Consonant gradation k:0

Consonant gradation l:0

Consonant gradation m:0

Consonant gradation n:0

Consonant gradation p:0

Consonant gradation s:0

Consonant gradation ŋ:0

Consonant gradation f:0

Consonant gradation r:0

Consonant gradation v:0

Consonant gradation j:0

Consonant gradation t:0

Gradation Series 4, II-I, tj and ts

Change rules

The Cx:Cy format was kept for hk:g, hp:b, ht:d, since the left context h:0 was unique, and no compilation conflict thus arose.

The bb:pp, gg:kk, dd:tt alternations were split into three rules, since keeping them in one Cx:Cy rule created compilation conflicts. Also, d:t contain a rule not found for the other two…

Gradation Series 4, II-I

bb:pp

gg:kk

g:k change for clitic -ge

dd:tt and dtj, dts

Gradation Series 7, III-II, ks(t), kt, ktj, kts

Exceptional II-III inverse gradation in present participles

This gradation is only for II-I syllable verbs that get III as present participles.

Candidates:

Strategy: Do insertion rule for the initial element.

Consonant insertion as II-III strengthening gradation with bm, gŋ

Consonant insertion as II-III strengthening gradation with dn/j + as I-III strengthening gradation with d

Consonant insertion as II-III strengthening gradation with hk, hp,

Consonant insertion as II-III strengthening gradation with htt(j/s)

Debugging of twol-rules

All rule conflicts have been successfully resolved. The rule file should be kept that way. Look out for conflicts in the compilation process, and resolve them as they appear!

This (part of) documentation was generated from src/fst/morphology/phonology.twolc