Finite state and Constraint Grammar based analysers, proofing tools and other resources
View the project on GitHub giellalt/lang-sme
The file format is documented in the Xerox manuals, especially in Karttunen 1993 Finite-State Lexicon Compiler, but see also the Beesley and Karttunen book. The file itself consists of a section defining Multichar_symbols, and of a large number of lexica, 183 lexica according to the present count (19.10.01). The file sme-lex.txt contains a.o. the continuation lexica for nouns, verbs and adjectives, whereas the bulk of the lexicon is divided into different files, as indicated below.
In the sme-lex.txt file, the Multichar_Symbols section contains all grammatical tags, and all multicharacter members of the alphabet (the latter set is taken from the grammar file).
The Root lexicon points to the lexica of the different parts of speech: (for each sublexicon there is a pointer to the relevant file containing the sublexicon)
LEXICON Root
NounRoot ; ! -> noun-sme-lex.txt
ProperNoun ; ! -> the file sme-lex.txt itself
AdjectiveRoot ; ! -> adj-sme-lex.txt
VerbRoot ; ! -> verb-sme-lex.txt
Pronoun ; ! -> closed-sme.lex.txt
Adverb ; ! -> adv-sme-lex.txt
Particles ; ! -> closed-sme.lex.txt
Subjunction ; ! -> closed-sme.lex.txt
Conjunction ; ! -> closed-sme.lex.txt
Adposition ; ! -> pp-sme.lex.txt
Postposition ; ! -> pp-sme.lex.txt
Preposition ; ! -> pp-sme.lex.txt
Interjection ; ! -> closed-sme.lex.txt
The different part of speech lexica are documented here, in the order just given.
The file contains the following sections:
The GuessNoun lexiconbelongs to the guess-sme binary file. The idea behind this is to have a binary file that guesses the grammatical form of nouns on the basis of their suffix, a hypothetical form like plineraiguin could for example be hypothetised to be the Comitative Plural of plinera. The guesser has not been used, and it is unclear whether it works at all. TODO: Check this.
The lexicon NounRoot first has some temporary sections. There is a set “compounds, awaiting a solution on the shorting question”, such like sámegiel, oar, beal, etc., reduced first- and middlepart compounds. The long-term solution is to build a syllable shortener for these compounds, for the time being the lexicalised ones are listed here. This section also contain some loanwords that typically have a shorter, non-sámi form as compounds, and a longer, Saami form when used as independent word. An illustrative example is the word sosial. In isolation, we find sosiála, but as compound, we find sosialdepartemeantta, etc. Thus, these words are listed here, pointing directly to the compound lexicon R.
Finally, the noun stem section itself is the large one, it contains about 26500 nouns, they are divided into different noun stem classes. They are documented in the next section.
The sublexica, ordered by inflectional type
Bisyllabic nouns:
GOAHTI !Bisyll. V-Nouns. Short nom-compound-forms goahte-,long/short
gen
GOAHTILONG !Long nom-compound-forms,long/short gen
GOAHTILONGSHORT !Sometimes long nom-compound-forms,long/short gen
ALBMI !Bisyll. V-Nouns. Short nom-compound-forms, long gen.
ALBMILONG !Bisyll. V-Nouns. Long nom-compound-forms, long gen.
ALBMILONGSHORT !Bisyll. V-Nouns. Long/SHORT nom-compound-forms, long
gen.
AIGI !Bisyll. V-Nouns. Short nom-compound-forms, short gen
ACTOR !Long compound-forms
ACTORLONGSHORT !Sometimes long compound-forms
ACTORTV !deverbal nouns from transitive verbs
ACTORLONGSHORTTV !deverbal nouns from transitive verbs
STAHTA !Bisyll. Non-Gradating a-Nouns; i-Illative
IIJA !Bisyll. Non-Gradating a-Nouns; a-Illative
LUONDU !this word (+vuohta) because of behavior in compounds
RUOKTU !only this word because of its behavior in compounds
MANNI !Bisyll. V-Nouns. Long/SHORT nom-compound-forms, long
gen. ILL:mánnii/mánnái
BASSI !Bisyll. V-Nouns. Short nom-compound-forms, short gen., long
heapmi-caritive
EADNI !Bisyll. V-Nouns. Short nom-compound-forms, long gen.short
caritive
RAFI !Bisyll. V-Nouns. Long nom-compound-forms, long gen. short
heapmi-caritive
DUOHKI !only this word so far for disamb. reasons
JOHTOLAT
LEXMUSH
Actios:
BOAHTALADDAN ! No compound-forms
BOAHTALADDANTV !Words coming from transitive verbs, No compound-forms
IHTALUDDAMAT ! No compound-forms, plural only
UPMI
EAPMI
MUITTASJEAPMI
EAPMITV !words stemming from transitives
MUITTASJEAPMITV !words stemming from transitives
Without loc and ill sg:
OLGU !Bisyll. V-Nouns. Short nom-compound-forms, short gen.
MIEHTI !Bisyll. V-Nouns. Short nom-compound-forms, long gen.
LULLI !Bisyll. V-Nouns. Long/SHORT nom-compound-forms, long gen
With comparatives:
GADDI !Bisyll. V-Nouns with Comparative Forms. Short nom-compound-forms,
long gen.
GADDILONG !long compound forms
GADDISHORT !Bisyll. V-Nouns with Comparative Forms. Short
nom-compound-forms, SHORT gen.
With comparatives, without loc and ill sg:
GADDILONGSHORT !NB! No SgIll and SgLoc because davvi is the only word
this far Bisyll. V-Nouns with Comp. Forms, long-short nomcmp, long
gencmpOARJI !Bisyll. V-Nouns with Comparative Forms. Short
nom-compound-forms, long gen.
LULLILONG !long compound forms
Plurals:
BUVSSAT !Pl.
BUVSSATLONG
BORALMASAT !like JOHTOLAT but plural only
Trisyllabic nouns:
Ending with consonant, gradating:
MATTAR !Tris. Anim. Gradating C-Nouns
MALIS !Tris. Inanim. Gradating C-Nouns, Short compound-forms
MALISLONG !Tris. Inanim. Gradating C-Nouns. Long compound-forms
MALISLONGSHORT !Tris. Inanim. Gradating C-Nouns. Long and short
compound-forms
OVCCIS ! Collective numerals
DAIVVAS !Tris. Gradating C-Nouns, The Troms declension: dáivvaš:dáivaha,
bearaš:bearraha, njunuš:njunnoha
Ending with vowel, gradating:
BEANA !Trisyll. Anim. Gradating 0-Nouns
SEAMU !Short compound-forms. Trisyll. Inanim. Gradating 0-Nouns
SEAMULONG !Long compound-forms. Trisyll. Inanim. Gradating 0-Nouns
Ending with consonant, non-gradating:
GAHPIR !Short compound-forms. Trisyll. Non-Gradating C-Nouns
GAHPIRLONGSHORT !Long and short compound-forms. Trisyll. Non-Gradating
C-Nouns
GAHPIRLONG !Long compound-forms. Trisyll. Non-Gradating C-Nouns
LEXDIMINC !Diminutives
Actios:
BOAHTINLONGSHORT ! Deverbal nouns
BOAHTIN ! Long compound-forms
BOAHTINSHORT ! Short compound-forms
BOAHTINLONGSHORTTV !!Words coming from transitive verbs Deverbal nouns
BOAHTINTV !!Words coming from transitive verbs Long compound-forms
BOAHTINSHORTTV !!Words coming from transitive verbs Short
compound-forms
IHTAMAT !Actio plurals
Plurals:
BADJOSAT !Pl. bajus:badjosat
BADJOSATLONG !Pl. bajus:badjosat
ALIMAT !Pl. alin:alimat
CEAKCAGAT !Like seamu but plural only
VUOIGNAHAT !Like DAIVVAS but only Pl. vuoiŋŋaš:vuoigŋahat
Contracted nouns:
BOAZU !Anim. Contracted 0-Nouns
SUOLU !Short compound-forms. Inanim. Contracted 0-Nouns
SUOLULONG !Long compound-forms. Inanim. Contracted 0-Nouns
GISTTA !The Noun gistta, gist -
FALIS !Contracted Anim. C-Nouns
LASIS !Contracted Inanim. C-Nouns
GUOVTTIS ! Collective numerals
GUOVTTU !Here because other lexicons don’t fit for it
Plurals:
DURVAT !like LASIS but pl. only
GIRKOSADDOT !LIKE SATTU but pl, only
Miscellaneous noun types:
GENTLEMAN ! cns-final bisyllabic loanwords (stem mana-
MADII !Another peculiar word that deserves its own lex
MASAI !heavy fin syll !gen = -a ill = -ii !parallel to NYSTØ
BUDEITA !Rather special word
GARGIA ! light fin syll, bisyll. on -o- that doesn’t have change o:u in
front of j (i): Kino
BUFFALO !heavy fin syll !gen = -a ill = -ii !parallel to NYSTØ
KULTUR !Recent loanwords on -vra with short cmp-form
KANTUR
MASKIN !Recent loanwords on -iidna with short cmp-form
TELEFON !Recent loanwords on -vdna with short cmp-form
SOSIAL !Recent loanwords on -ála with both short and long cmp-form
OLLUVUOHTA !Exceptional vuohta-Nouns
MAGASH
EGEZHAGAT
tbw. Here we document the nominal section of the sme-lex.txt file.
The proper nouns are stored in gt/sme/src/propernoun-sme-lex.txt.
Propernoun is one file, with the inflectional lexica first, followed by the over 36000 items long proper noun lexicon. This lexicon points to several sublexica. They are shown below, ordered according to the phonological and semantical properties of the stem. After the table comes a list of lexica not yet integrated in the table.
Type
StemCoda CG IllChange Loc Lexicon
--------------------------------------------------------
Monosyllabic stems
HeavyVow no no -as NYSTØ
HeavyCns no no -as BERN
Bisyllabic stems
LightVow no yes -s ACCRA
LightVow yes yes -s MARJA
LightVow yes yes -s SUOPMA
Light e no no -s SIJTE
LightCns no HEANDARAT
LightVow no no -s NIKOSIIJA
LightVow yes yes -s HEIM
LightVow yes yes -s SUND
Trisyllabic stems
LightCns no no -is LONDON
LightCns no no -is ANAR
LightCns yes yes -is DUORTNUS
LightVow yes GUOLBBA
LightCns no no -is PLACE-DIM
LightCns yes yes -is RANES
LightCns yes yes -is CAVKKUS
LightCns no no -is BALAK
LightCns no no -is SARAK
Contracted stems
DAVVISUOLU !Inanim. Contracted 0-Names - from SUOLU
GEAVNNIS
Mixed stems
-nen no no -as/nenis no C-FI-NEN
HAWAII
SKOHTERMADII
Plural stems
ALEUHTAT is for bisyllabic, would-have-been vowel final stems
VARGGAT is as aleuhtat, but with Sg Gen and Loc substandard forms
EATNAMAT is for trisyllabic plural stems
ADJAGAT is for trisyllabic plural stems
SULLOT is for contracted plural stems
LASSAT is for contracted plural stems
SKANIK is for plurals on -k
---------------------------------------------------------
Notes:
Comp means "has comparative forms"
This table is still not completed. The lexica are there, but they are not all properly characterised.
Looking at the distribution, and ignoring the semantic subtags, we have a (slightly outdated) distribution of names over lexicon types as follows:
9114 ACCRA
8482 BERN
8371 MARJA
6480 LONDON
1390 C-FI-NEN
865 NYSTØ
296 PLACE
133 DUORTNUS
98 DAVVISUOLU
68 NIILLAS
62 GUOLBBA
59 ANAR
48 ALEUHTAT
31 SULLOT
28 HEANDARAT
27 GIEDDI
13 VARGGAT
12 GEAVNNIS
7 EATNAMAT
Saami names have been added from Norway, in 2002. Since that, Statens Kartverk has translated most (all?) Saami names from the old Bergsland/Ruong to the new 1979 orthography. Now, the remaining Sámi names should be added to the list, in the following way:
Saami names from Finland and Sweden have not been added, but they are underway.
Saami names from outside Sápmi, as listed in Sámi Atlas are added. Other Saami names from outside Sápmi should also be added, but we don’t know of good sources for such names (language councils are one possibility)
Names in other Saami languages are increasingly being used in North Saami texts, e.g. when referring to South Saami or Lule Saami institutions. These ones should be added and organised in lexica.
The file adj-sme-lex.txt contains 4270 entries.
In the lexicon file adj-sme-lex.txt, the sublexica are distributed in the following way (30.06.05) (ordered after frequency, thereafter after declension type):
Making linguistic sense of the system (Sammallahti's codes aaa etc.):
899 DABALAS aad
477 JEAGOHEAPMI bae
358 BOAKKAS c
329 At attributes
280 BEAKKAN aab
218 GAPPUS aab
205 JOHTIL babaa
191 NUORRA aaa
165 AKTIIVA
157 JIEDNAI bad
150 GARAS bbb
139 MEAHTTUS aab
114 LAIKI baa
106 DEARVVASLAS aad
71 ISSORAS aad
59 VIELG babab
38 CAHKK babaa babab
31 BUOREMUS aae
30 JUHKKIS aaa
26 GUOHCA aab
26 EATTAS babba
19 LODJI baa bac
14 GEARGGUS aab
13 SEARRA baa
13 DILDDAS babba
8 VUDDJII bad
8 CIENAL babbb
7 LINIS bbe
7 DEAHTIS bbe
7 BIEKKUS babba
7 ASEHAS baf
6 OVDDIT aae
6 NUOLUS aab
5 VATTIS aab
5 NJUORAS bbb
5 HEAHKAS babba
4 VIISSIS aab aac
4 LIEKKUS aab
4 FINJU-
3 UHCC bba
3 SUVRRIS bbe
3 JALGAT bbc babab
3 BUORRE ab
3 BU/MUS
3 ATTR
3 ALLAT bbc
2 NJALGGAT bbe
2 NAMAT
a.................No morphologically distinct attr. form
aa..............attr. is not inflected
aaa...........Bisyll.
NUORRA
JUHKKIS Non-gradating
aab...........Trisyll.
BEAKKAN Gradating
GAPPUS Gradating
MEAHTTUS Compounded non-gradating
GUOHCA Gradating
GEARGGUS Gradating
NUOLUS Gradating
VATTIS Gradating
VIISSIS Gradating
LIEKKUS Gradating
IPMAHA Gradating
aac...........Contracted
VIISSIS
aad...........Qadrisyll.
DABALAS
DEARVVASLAS
ISSORAS
aae...........Comparative forms
BUOREMUS
OVDDIT
ab..............Attr. in partial congruence with the noun
BUORRE
b.................Morphologically distinct attr. form
ba..............Attr. form ends with -s
baa...........Bisyll.
LAIKI Attr. form -es in WeG
LODJI Attr. form -es and -is in WeG
SEARRA Attr. form -s in WeG
bab...........Trisyll.
baba........Non-gradating
babaa.....No contraction in attr. form
JOHTIL is-Attr.
CAHKK is-Attr.
babab.....Contraction (and ensuing strengthening) in attr. fom
VIELG es-Attr.
CAHKK es-Attr.
JALGAT es-Attr.
babb........Grade alternation in stem
babba.....s-stems with WeG attr-form
EATTAS is-Attr.
DILDDAS is-Attr.
BIEKKUS is-Attr
HEAHKAS is-Attr
babbb.....Attr. form in StG
CIENAL is-Attr.
bac...........Vocalic contracetd attr. form
LODJI -es and -is Attr. in WeG
bad...........Contracted stems
JIEDNAI Non-gradating
VUDDJII Non-gradating
bae...........Caritives with attr- -his.
JEAGOHEAPMI
baf...........Qadrisyll. Adj., with Attr.-his
ASEHAS
bb..............Attr. form ends with -a
bba...........Bisyll. ending in -i (or -a sometimes)
UHCC StG attr.
bbb...........Trisyll. ending -as
GARAS Gradating with StG attr.
NJUORAS Gradating with StG attr.
bbc...........Trisyll. ending -at
JALGAT Non-gradating with StG attr.
ALLAT Non-gradating with StG attr.
bbe...........Trisyll. ending -is
LINIS Gradating with StG Attr.
DEAHTIS Contracted with StG Attr.
SUVRRIS Gradating with WeG Attr.
NJALGGAT Gradating with StG Attr.
c.................No attr.
BOAKKAS Trisyll.
The lexicon is stored in the verb-sme.txt file.
VerbRoot contains 49 sublexica divided into three stem types:
Modals:
**
GALGA
FERTE
Bisyllabic verbs:
Impersonals:
ARVI arvit sataa !Bisyllabic Impersonal Verbs
ARVALADDAT arváladdat sataa !Already derived bisyllabic Impersonal
Verbs
Transitives:
DIEHTI diehtit tietaa !Bisyllabic i-Verbs with Personal Passive
BORRA !Bisyllabic a- and u-verbs with Personal Passive
DEAKCU !as BORRA for u-verbs with dim -astit, and a-verbs with dim
-istit that are hardcoded
DIEHTISHORT !Short actio compound-form DIEHTILONGSHORT !Long and short
actio compound-form DIEHTALADDAT diehtáladdat tietaa !Already derived
bisyllabic Verbs with Personal Passive
HAHTTIT !Four-syll kausatives on -httit
DAHTU dáhtut ! As diehti, but -ut verbs, thus without short passive
BOLTU ! As dáhtu but with dim -astit that are harcoded
DOAROSTUVVA !INCHOATIVES on -stuvvat
Intransitives:
ALLU ! -ut verbs, thus without short passive BOAHTI boahtit tulla
!Bisyllabic i-Verbs without Personal Passive
BOAHTILONGSHORT !!Long and short actio compound-form DIEVVA ! Bisyllabic
a- and u-verbs without Personal Passive but with Actor BOAZZU !as DIEVVA
for u-verbs with dim -astit, that are hardcoded BINDU !as DIEVVA (but
without short passive) for u-verbs with dim -astit, that are hardcoded
BOAHTALADDA boahtáladda tulla !Already derived bisyllabic Verbs without
Personal Passive
RAIMMAHALLA !passives on -hallat and INCHOATIVES on -stuvvat
Personal Passive but with Acc obj:
MAHTI máhttiit ! Bisyllabic Verbs without Personal Passive but with Acc
obj.
MAHTALADDA máhtáladdat ! Already derived bisyllabic Verbs without
Personal Passive but with Acc obj.
Inherent passives:
RAIMMAHALLA !passives on -hallat and INCHOATIVES on -stuvvat UVVA
!passives -uvvat
Contracted verbs:
Impersonals:
BORGE borget tehda pyry !Contracted Impersonal Verbs
Transitives:
DOHPPE dohppet tarttua !Contracted Verbs with Personal Passive
MUITA ! Inchoatives and translatives on -á, -o, -e with Personal
Passive
Intransitives:
GILLE gillet viitsia !Contracted Verbs without Personal Passive
CIRRO ! Inchoatives and translatives on -á, -o, -e without Personal
Passive
Personal Passive but with Acc obj:
MAHTA máhttát !Contracted Verbs without Personal Passive but with Acc
obj.
Inherent passives:
BASSO bassot ! Bisyllabic, inherently passive -ot verbs
Trisyllabic verbs:
Impersonals:
COASKKIT čoaskkidiit !Trisyllabic impersonal verbs
ARVVASJ arvvašit !Trisyllabic impersonal verbs ending -šit, -skit, smit,
-idit, -ldit, -git and 5-syllables
ARVIL arvilit !Trisyllabic impersonal verbs ending -lit
Transitives:
MUITAL muitalit !Trisyllabic Verbs with Personal Passive
MUITTASJ !Words ending -šit, -skit, -ldit - !directed here as well:
!Reciprocals on -dit !Momentatives on -dit, -ádit, -ihit, -e7hit
!Frequentatives on -(u)hit !Continuatives on -nit !Inchoatives in -nit
HALIID !Words ending -smit, -idit, -git - BONJAT !!Cont/Freq on -dit,
Continuatives on -(u)hit, Reciprocals, momentatives and frequentatives
ending -alit
VUORDIL !Trisyllabic Verbs ending -lit, -rit with Personal Passive
Intransitives:
ALIST alistit !Trisyllabic Verbs without Personal Passive
BEAGASJ !Words ending -šit, -skit -ldit, transitives on -hit
!directed here as well: !Reciprocals on -dit !Momentatives on -dit,
-ádit, -ihit, -e7hit !Frequentatives on -(u)hit !Continuatives on -nit
!Inchoatives in -nit
JORGGIID !Words ending -smit, -idit, -git
BALAT !!Cont/Freq on -dit, Continuatives on -(u)hit, Reciprocals,
momentatives and frequentatives ending -alit
SUOTNJAL suotnjalit !Trisyllabic Verbs without Personal Passive ending
-lit
BOTNJAS botnjasit !Trisyllabic Verbs without Personal Passive ending
-nit and -sit
LASSAN !Trisyllabic Verbs ending -nit without Personal Passive
Personal Passive but with Acc obj:
GEAGAT ! Trisyllabic Verbs without Personal Passive but with Acc obj.
BUOVVAL buovvalit ! Trisyllabic Verbs without Personal Passive but with
Acc obj ending -lit
The stems are distributed numerically as follows (the -it class includes both even-syllable and odd-syllable verbs):
-at 3722
even-syll -it 1035
-ut 794
total 5551
3-syllabic -it 5376
-át 297
-et 2310
-ot 111
total 2718
tbw.
Here documenting the main even-syll ones, the other ones are similar. DIEHTI is transitive, BOAHTI is intransitive.
DIEHTI -> +V: DIEHTIStem ; +V: DeverbalVerbsDIEHTI ;
BOAHTI -> +V: BOAHTIStem ; +V: DeverbalVerbsBOAHTI ;
DIEHTIStem -> :Y7j PASSIVE ; BOAHTIINCH ;
BOAHTIStem -> SG3PASSV ; BOAHTIINCH ;
BOAHTIINCH -> DeverbalNounsV ; +goah0ti:goah'ti BOAHTICnj ; BOAHTICnj ;
BOAHTICnj -> +Ind+Prs: PrsV ; +Ind+Prt: PrtV ; +Pot+Prs:Q7z1 PrsC ;
+Cond: CondV1 ; +Imprt: ImprtVA ; NominalFormsV ;
NominalFormsV -> :X1 NominalFormsV1 ; :X4 NominalFormsV2 ;
:Q6 NominalFormsV3 ; :X2 NominalFormsV4 ; :Q3 NominalFormsV5 ;
:Y1 NominalFormsV6 ;
PASSIVE -> +Pass:uvvo DOHPPEINCH ; +Pass+meahttun+A:uvvomeahttum
MEAHTTUN ; +Pass+PrfPrc:un K ; +Pass+eaddji+N+Actor:uvvojeaddji¤
DEVNVCASE ; +Pass+upmi+N:upmi DEVNVCASE ;
DeverbalVerbsDIEHTI ->
+st:X8st MUITALStem ;
+st+alla:X6stalla DIEHTIStem ;
+st+adda:X6stadda DIEHTIStem ;
+l:l MUITALStem ;
+l+adda:X2ladda DIEHTIStem ;
+l+ahtti:lahtti DIEHTIStem ;
+l+asti:las'ti DIEHTIStem ;
+h:X4h MUITALStem ;
+h+alla:X6halla DIEHTIStem ;
+h+adda:X6hadda DIEHTIStem ;
+h+asti:X4has'ti DIEHTIStem ;
+stuvva:X8stuvva SG3PASSV ;
+d:Q8d MUITALStem ;
DeverbalVerbsBOAHTI ->
+st:X8st ALISTStem ;
+st+alla:X6stalla BOAHTIStem ;
+st+adda:X6stadda BOAHTIStem ;
+l:l ALISTStem ;
+l+adda:X2ladda BOAHTIStem ;
+l+ahtti:lahtti BOAHTIStem ;
+l+asti:las'ti BOAHTIStem ;
+h:X4h MUITALStem ;
+h+alla:X6halla DIEHTIStem ;
+h+adda:X6hadda DIEHTIStem ;
+h+asti:X4has'ti DIEHTIStem ;
+stuvva:X8stuvva SG3PASSV ;
+d:Q8d ALISTStem ;
Within each of the main groups, there are FIVE types, impersonal verbs, verbs with and without personal passives, verbs without Personal Passive but with Acc obj (+ two more lexicas, see above The VerbRoot lexicon), and inherent passives. The difference between i/a/u and e/á/o verbs is handeled in the rules file, and not in the lexicon file.
The with / without Personal Passive distinction shows up in one sublexicon. DOHPPE has PASSIVE, where GILLE has SG3PASS. So, this is (probably) a transivity difference, cf. also diehtit vs. boahtit. It seems thus that the difference is one of transitivity: 0, 1 and 2 valence.
At present, the file verb-sme-lex.txt comtains all the verbs. In the beginning of the file, all sublexica are exemplified. Then follows the bulk of the verbs, twosyllabic even, manysyllabic even, odd and contracted verbs.
The tag system follows the outline in Nickel.
All Pronouns have the initial lexicon path Root -> Pronoun -> …
Lexicon path:
Personal
firstperspron
firstperspronsg -> wordforms -> K
firstpersprondu -> wordforms -> K
perspronpl -> wordforms -> K
nonfirstperspron
nonfirsperspronsg -> wordforms -> K
nonfirstpersrondu -> wordforms -> K
perspronpl -> wordforms -> K
Note that 3rd person is identical for all three persons. Not all forms were different for the sg and du forms, but the lexica were split for consistency.
Mii, Gii, Guhte, Guhtemuš, Makkár, MAn Láhkái. The sublexicon Interrogative contains one entry for Sg Nom, and points the rest to the case paradigms.
Interrogative +Sg+Nom -> K (one entry for gii and one for mii)
oblintercas (one entry for gii and one for mii) demcas
The lexicon path:
Demonstrative
demcas (one entry for each stem)
demcassg
nomdemcassg -> wordforms -> K
obldemcassg -> wordforms -> K
demcaspl
nomdemcaspl -> wordforms -> K
obldemcaspl -> wordforms -> K
The Nominative forms are just listed. The oblique ones are directed to the sublexicon reflobl, and there directed via different case stems to appropriate Px sublexica. These sublexica are the same as the ones for nouns, they are found in the sme-lex.txt file. The only exception are some sublexica that are used only for plural forms, these are duplicated here from the sme-lex file, in order not to revise the main lexicon.
The section on reciprocal pronouns consists of three parts. The first 6 entries handle the first element of the recipr. The next 12 handle the 2nd part of the non-Px recipr. Finally, the members of the third section point to special Px sublexica, designed for the reciprocal pronouns, and found in the same section.
Formally, the relative and interrogative pronouns are identical. In this parser, we skip the separate chapter on relative pronouns, and instead we use the interrogative pronouns.
We divide the indefinite pronouns into three groups, with a fourth group of leftovers waiting for a better destiny:
These pronouns have two stems, one nominative, and one oblique, and the
clitic -ge attached to the case ending. The initial lexicon splits
them in two, one hard-coded nominative (e.g. giige), and one oblique
stem (e.g. gea-). Then, the case + clitic sequence is treated as a
single suffix (e.g. locative -sge, etc.). Since the clitic slot has
already been filled, they are directed to # rather than to K.
This section hosts a seemingly complicated system of taylored sublexica. It contains three sections: First a section where the pronouns themselved are split into different continuation lexica, then a section with intoermadiate lexica, and finally a section with the case suffixes themselves. The lexica are partly modelled upon nominal lexica.
Naming convention for the sublexica:
There is first a list of multiword indefinites. These are picked out by the preprocessor and copied onto a file abbr.txt and put in the bin/ catalogue. In thie closed-sme-lex.txt they behave like the other indeclineable indefinites.
Indefinite pronouns are complex, and the grammars are not always explicit enough, so this section hosts a set of pronouns, partly with a hard-coded tag, partly just commented out. They are awaiting a principled linguistic solution, but in order to do that, we need more info than we can get from the reference grammars.
TODO: Have a linguistic/native speaker-look at this section.
The numeral lexica are formed as a generator, generating all possible
numerals. The basic lexicon is Numeral, and it looks like this:
LEXICON Numeral
MILJON ; ! a noun of its own
UNDERDUHAT ; ! for generator under 1000
JUSTDUHAT ; ! going via 1000
OVERDUHAT ; ! for generator over 1000
OLD ; ! for "thirteen hundred, etc.
!num-basic ; ! replaced by the 5 lexica above
!num-derived ; ! still unimplemented
num-imprecise ;! still almost unimplemented
ARABIC ; ! for the arabic numerals
ROMAN ; ! for the roman numerals
MILJON is a noun. OLD is the old way of counting. num-ordinal act like adjectives, they are not finished yet. ARABIC and ROMAN contain number generators.
So, what is the reason for the three different lexica around 1000?
The reason is that the numeric system turns at the thousand mark. Numbers above it and numbers below it behave in the same way, thus we have both twentyfour and twentyfourthousand, etc.
The path is OVERDUHAT -> JUSTDUHAT -> UNDERDUHAT. OVERDUHAT generates the part of the numeral that is over 1000, and all these lexica then point to JUSTDUHAT. That lexicon has an optional “(one) thousand” before it leads either to DUHAT and via the relevant case paradigm to K, or to UNDERDUHAT. UNDERDUHAT contains the numerals 1-999. UNDERDUHAT starts with the lexicon for one, and gives each group of numerals its own lexicon.
The cardinal and ordinal numbers are split at the final lexica, the
OKTAF and 2TO9F lexica. This generates both numbers as second and
fiftysecond.
All the lexica for indeclinable words are made the same way:
LEXICON Root
Adverb ;
LEXICON Adverb
áđamusat adv ;
LEXICON adv
+Adv:0 K ;
The Root lexicon points to the POS lexica (Adverb etc.). Each of the POS lexica lists the entries, with a pointer to an arbitrarily named sublexicon (here “adv”). This sublexicon contains the grammatical tag for the POS in question (the tag has no surface form, hence “:0”), and eventually a pointer towards the cliticon lexicon K. Adverbs can have clitics added, hence K, whereas subjunctions do not, hence no K.
[XXX At the moment particles are not directed to K, perhaps they should be. TODO: Check with corpus and native speakers.]
They are explained in the intro to the section “Indeclinable words” above.
These are in the closed-sme-lex.txt file. Their tag is +Pcle and th|qeir lexicon path is:
Root -> Particle -> pcle -> #
Subjunctions are ahte, juos, etc. These are in the closed-sme-lex.txt file. Their lexicon path is:
Root -> Subjunction -> -> #
Conjunctions are ja, dahjege, etc. These are in the closed-sme-lex.txt file. Their tag is +CC and their lexicon path is:
Root -> Conjunction -> Cc -> #
There are three different classes here: Postpositions, occuring after their complement, prepositions, occuring before, and adpositions, occuring both before and after. This could have been done the Lingsoft way as well: Having +Adp as a common tag for both, and eventually +Prep and +Postp as subtags, no subtag would indicate both ways (or both subtags could be used). At the moment, they are left as 3 distinct groups. The classification is based upon Nickel, p-positions found only in Sammallahti’s dictionary and not in Nickel were put in the Adposition group. Empirical studies will probably lead to rearrangement of the present division, this should be looked into in connection with the morphological disambiguator (cg grammar).
Adpositions are are bajil, birra, gaskal, etc. These are in the pp-sme-lex.txt file. Their tag is +Adp and their lexicon path is:
Root -> Adposition -> Pp -> #
Postpositions are are bokte, lusa, etc. These are in the pp-sme-lex.txt file. Their tag is +Po and their lexicon path is:
Root -> Postposition -> Postp -> #
Prepositions are are aisttan, earet, etc. These are in the pp-sme-lex.txt file. Their tag is +Pr and their lexicon path is:
Root -> Preposition -> Prep -> #
Interjections are are hoi, huh, kyš-kyš, etc. These are in the closed-sme-lex.txt file. Their tag is +Interj and their lexicon path is:
Root -> Interjection -> Ij -> #
There is a file called abbr-sme-lex.txt.