irish lfg grammar development - uni...

Irish LFG Grammar Development

Sebastian Sulger

Universitat Konstanz

August 28, 2008

Sebastian Sulger (Universitat Konstanz) Irish Grammar Development August 28, 2008 1 / 24

Preview

1 General remarks on Irish

2 Two phenomena of Irish Morpho-SyntaxLenitionEclipse

3 Irish Morphology in xfstVerbsNouns

4 Morphology - Syntax Interface

5 LFG Grammar


General remarks on Irish

Irish - language history

among the oldest languages of Europe





fragments (inscriptions in stones) date from the 4th to the 6th century






related to Scottish Gaelic and Manx (which is regarded as extinct)






related to Scottish Gaelic and Manx (which is regarded as extinct)

Indo-Germanic > Celtic > Insular Celtic > Goidelic (”Gaelic”) > Irish(Scottish Gaelic, Manx)



Irish today

first official language of Ireland



Irish today


minority language; spoken predominantly in the so-called Gaeltachtaı

(population: 91,862, 60% thereof: Irish in everyday life)



Irish today


minority language; spoken predominantly in the so-called Gaeltachtaı

(population: 91,862, 60% thereof: Irish in everyday life)

1,8 million people (one out of three) on the island speak Irish ”to acertain degree”



Dispersion of the Gaeltachtaı

Figure: Map of Ireland including Gaeltachtaı



Irish today: standard language

fear: Irish could be endangered by the discrepancies between thedialects -> creation of the standard form An Caighdean Ofigiuil

(”The Official Standard”, OS)






combination of syntactic and lexical elements of all dialects







pronunciation is based on the Connacht dialect








my implementation uses this standard form








my implementation uses this standard form

OS Irish: Conas ata tu? ’How are you?’


Two phenomena of Irish Morpho-Syntax Lenition

Lenition

Initial Mutation



Lenition

Initial Mutation

a somewhat ”strong” consonant is ”weakened” (lat. lenis ’soft’)



Lenition

Initial Mutation

a somewhat ”strong” consonant is ”weakened” (lat. lenis ’soft’)

affected consonants:b (/b/) -> bh (/v/)

c (/k/) -> ch (/>kx/)

d (/d/) -> dh (/g/)f (/f/) -> fh (ø)g (/g/) -> gh (/È/) -> orthography: h

m (/m/) -> mh (/v/)p (/p/) -> ph (/f/)s (/s/) -> sh (/h/)t (/t/) -> th (/h/)



Lenition

consonants that are not affected: h, l, n, r



Lenition


lenition is realized in various lexical and syntactic contexts



Lenition



lenition needs an external trigger in front of the lenited consonant(exception: verbs in the preterite, imperfect and conditional)



Lenition



lenition needs an external trigger in front of the lenited consonant(exception: verbs in the preterite, imperfect and conditional)

example: prefix an-, used for intensification of adjectivesbeag /b/ ’small’; an-bheag /v/ ’very small’te /t/ ’hot’; an-the /h/ ’very hot’


Two phenomena of Irish Morpho-Syntax Eclipse

Eclipse

Initial Mutation



Eclipse

Initial Mutation

voiceless plosives and /f/ become voiced; voiced plosives becomenasalized



Eclipse

Initial Mutation


eclipse is also possible with vowels -> n- is added in front of the vowel



Eclipse

Initial Mutation


eclipse is also possible with vowels -> n- is added in front of the vowel

affected consonants:b (/b/) -> mb (/m/)c (/k/) -> gc (/g/)d (/d/) -> nd (/n/)f (/f/) -> bhf (/v/) -> orthography: letters for newg (/g/) -> ng (/ñ/) pronunciationp (/p/) -> bp (/b/)t (/t/) -> dt (/d/)



Eclipse

consonants that are not affected: h, l, m, n, r, s



Eclipse


eclipse is realized in various lexical and syntactic contexts



Eclipse



eclipse needs an external trigger in front of the eclipsed consonant (noexceptions spotted so far)



Eclipse



eclipse needs an external trigger in front of the eclipsed consonant (noexceptions spotted so far)

example: prepositional phrases containing the preposition iteach (/t/) ’house’; i dteach (/d/) ’in a house’Corcaigh (/k/) ’Cork’; i gCorcaigh (/g/) ’in Cork’


Irish Morphology in xfst

xfst System

14 adjectives, 15 verbs, 71 nouns, 15 (inflecting) prepositions



xfst System


16 phonological rules written so far to account for:



xfst System



intensification of adjectives (prefixation of an + lenition)



xfst System



intensification of adjectives (prefixation of an + lenition)irregular plural forms of nouns



xfst System



intensification of adjectives (prefixation of an + lenition)irregular plural forms of nouns

multiword transducer to handle phonological processes across wordboundaries (Initial Mutations)


Irish Morphology in xfst Verbs

Verbs

15 verbs included (the auxiliary ta being one of them)



Verbs


verbs may be formed synthetically or analytically



Verbs



synthetical verb forms: person/number expressed in the endingtaim ’I am’; *taim me

molaim ’I praise’; *molaim me



Verbs





analytical verb forms: person/number have to be expressed by aseparate (pro)nounta siad ’they are’molann sibh ’you (pl.) praise’



Verbs






problems for syntax: verb and subject in a single word; sometimessynthetical, sometimes analytical, even within one paradigm



Verbs






problems for syntax: verb and subject in a single word; sometimessynthetical, sometimes analytical, even within one paradigm

cf. Bresnan (2001): optional use of the subject pronoun in paradigms


Irish Morphology in xfst Nouns

Nouns

71 nouns included



Nouns

71 nouns included

five declinations, each with several sub-groups



Nouns

71 nouns included


declination classes and sub-groups differ in forming the genitive andplural forms (among other differences)



Nouns

71 nouns included



approach: multiple lexicons for declinations and sub-groups



Nouns

71 nouns included




nouns are arranged in these lexicons and labeled accordingly; labelstrigger phonological rules to form the correct plural/genitive



Nouns

71 nouns included




nouns are arranged in these lexicons and labeled accordingly; labelstrigger phonological rules to form the correct plural/genitive

works: implemented for the nominative plural


Morphology - Syntax Interface


problem: (lexical) triggering of Initial Mutations





Initial Mutations operate across word boundaries






approach (cf. Mittendorf & Sadler (2006)):







list words in lexicon with all mutation variants (radical, lenition,eclipse), tagged accordingly







list words in lexicon with all mutation variants (radical, lenition,eclipse), tagged accordinglylist triggers in lexicon with governed mutation (e.g. preposition i :+Ecl+ for Eclipse)







list words in lexicon with all mutation variants (radical, lenition,eclipse), tagged accordinglylist triggers in lexicon with governed mutation (e.g. preposition i :+Ecl+ for Eclipse)set up multiword transducer that ensures that lexical mutationconditions are satisfied by the input string




Example: i ’in’, teach ’house’

xfst[1]: up i xfst[1]: up teach+Rad+i+Prep+PrepNoP+Ecl+ +Rad+teach+Noun+3P+Sg+Masc

CS 1: P:23

MUT_SFX_BASE:16

+Rad+:17

P_BASE:10

i:9

P_SFX_BASE:6

+Prep:7

PINFL_SFX_BASE:4

+PrepNoP:5

MUT_SFX_BASE:2

+Ecl+:3

CS 2: N:27

MUT_SFX_BASE:18

+Rad+:19

N_BASE:14

teach:11

N_SFX_BASE:8

+Noun:9

PERS_SFX_BASE:6

+3P:7

NUM_SFX_BASE:4

+Sg:5

GEND_SFX_BASE:2

+Masc:3




Example: i dteach ’in a house’ - resulting tree

xfst[1]: up i xfst[1]: up dteach+Rad+i+Prep+PrepNoP+Ecl+ +Ecl+teach+Noun+3P+Sg+Masc

CS 2: PP:67

P:58

MUT_SFX_BASE:17

+Rad+:18

P_BASE:11

i:10

P_SFX_BASE:7

+Prep:8

PINFL_SFX_BASE:5

+PrepNoP:6

TMUT_SFX_BASE:20

+T-Ecl:21

N:66

N_BASE:35

teach:32

N_SFX_BASE:29

+Noun:30

PERS_SFX_BASE:27

+3P:28

NUM_SFX_BASE:25

+Sg:26

GEND_SFX_BASE:23

+Masc:24




problem: synthetical verb forms





paradigm ta (directly from source file):ta+Verb+Pres+Pron1Sg:taim #;ta+Verb+Pres:ta #;ta+Verb+Pres:ta #;ta+Verb+Pres+Pron1Pl:taimıd #;ta+Verb+Pres:ta #;ta+Verb+Pres:ta #;





paradigm ta (directly from source file):ta+Verb+Pres+Pron1Sg:taim #;ta+Verb+Pres:ta #;ta+Verb+Pres:ta #;ta+Verb+Pres+Pron1Pl:taimıd #;ta+Verb+Pres:ta #;ta+Verb+Pres:ta #;

’taim’ and ’taimıd’ are synthetical verb forms -> labeling




approach: labels trigger templates in LFG grammar





templates assign the appropriate grammatical features to the subject






example:ta+Verb+Pres+Pron1Sg:taim #;






example:ta+Verb+Pres+Pron1Sg:taim #;

label ’+Pron1Sg’ activates template ’PronSFX1Sg’ in LFG grammar;template assigns the following to the subject:(ˆ SUBJ PRED)=’pro’(ˆ SUBJ CASE)=nom(ˆ SUBJ NUM)=sg(ˆ SUBJ PERS)=1(ˆ SUBJ PRON-TYPE)=pers.


LFG Grammar

LFG Grammar

grammar is yet very small: ”Grammar has 14 rules with 70 states, 77arcs, and 77 disjuncts (77 DNF).”


LFG Grammar

LFG Grammar


coverage: agreement, intransitives, transitives, passive construction(without agent)


LFG Grammar

LFG Grammar


coverage: agreement, intransitives, transitives, passive construction(without agent)

grammar is constantly under construction


LFG Grammar

Some C- and F-Structures

Example: taim fuar. ’I am cold.’

CS 1: ROOT:64

S:62

AUXP:61

AUX:48

táim:1

AP:60

A:59

fuar:16

PUNCT:36

.:35

"táim fuar."

'tá<[1-SUBJ:pro]>'PRED

'pro'PREDCASE nom, NUM sg, PERS 1, PRON-TYPE pers

SUBJ

CLAUSE-TYPE decl, TENSE pres

64626148

16059163635


LFG Grammar


Example: chaith sı ull. ’She threw an apple.’

CS 1: ROOT:66

S:64

V:46

chaith:1

NP:52

PRON:18

s�:17

NP:63

N:62�ll:19

PUNCT:37

.:36

"chaith sí �

ll."

'caith<[17:pro], [19:�

ll]>'PRED

'pro'PREDCASE nom, GEND fem, NUM sg, PERS 3, PRON-TYPE pers

521817

SUBJ

'�

ll'PREDCASE acc, GEND masc, NUM sg, PERS 3

636219

OBJ

CLAUSE-TYPE decl, TENSE past666446

13736


LFG Grammar


Example: nı chaith sı ull. ’She does not throw an apple.’

CS 1: ROOT:144

S:142

NEG:119

n�:25

V:125

chaith:28

NP:130

PRON:46

s �:45

NP:141

N:140�ll:47

PUNCT:66

.:65

"ní chaith sí �

ll."

'caith<[45:pro], [47:

ll]>'PRED

'pro'PREDCASE nom, GEND fem, NUM sg, PERS 3, PRON-TYPE pers

1304645

SUBJ

'

ll'PREDCASE acc, GEND masc, NUM sg, PERS 3

141140

47OBJ

'n �'PREDnegADJUNCT-TYPE119

25ADJUNCT

CLAUSE-TYPE decl, TENSE pres, VFORM neg144142125

286665


LFG Grammar


Example: staideareann me an chlairseach. ’I study the harp.’

CS 1: ROOT:70

S:68

V:49

staid�

areann:1

NP:54

PRON:15

m

:14

NP:67

D:17

an:16

N:66

chláirseach:18

PUNCT:37

.:36

"staidéareann mé an chláirseach."

'staid�

ar<[14:pro], [16:cláirseach]>'PRED

'pro'PREDCASE nom, NUM sg, PERS 1, PRON-TYPE pers

541514

SUBJ

'cláirseach'PREDCASE acc, DEF +, GEND fem, NUM sg, PERS 3

6717166618

OBJ

CLAUSE-TYPE decl, TENSE pres706849

13736


LFG Grammar

Thank you!


irish lfg grammar development - uni...

Documents