Lithuanian NLP Grammar

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

View the project on GitHub giellalt/lang-lit

Lithuanian language model documentation

All doc-comment documentation in one large file.

src-cg3-functions.cg3.md

These sets model noun phrases (NPs). The idea is to first define whatever can occur in front of the head of the NP, and thereafter negate that with the expression WORD - premodifiers.

The set NOT-NPMOD is used to find barriers between NPs. Typical usage: … (*1 N BARRIER NPT-NPMOD) … meaning: Scan to the first noun, ignoring anything that can be part of the noun phrase of that noun (i.e., “scan to the next NP head”)

These were the set types.

HABITIVE MAPPING

sma object

SUBJ MAPPING - leftovers

OBJ MAPPING - leftovers

HNOUN MAPPING

This (part of) documentation was generated from src/cg3/functions.cg3

src-fst-morphology-affixes-adjectives.lexc.md

Adjective inflection The Lithuanian language adjectives compare.

This (part of) documentation was generated from src/fst/morphology/affixes/adjectives.lexc

src-fst-morphology-affixes-nouns.lexc.md

Noun inflection The Lithuanian language nouns inflect in cases.

This (part of) documentation was generated from src/fst/morphology/affixes/nouns.lexc

src-fst-morphology-affixes-propernouns.lexc.md

Proper noun inflection The Lithuanian language proper nouns inflect in the same cases as regular nouns, but with a colon (‘:’) as separator.

This (part of) documentation was generated from src/fst/morphology/affixes/propernouns.lexc

src-fst-morphology-affixes-symbols.lexc.md

Symbol affixes

This (part of) documentation was generated from src/fst/morphology/affixes/symbols.lexc

src-fst-morphology-affixes-verbs.lexc.md

Verb inflection The Lithuanian language verbs inflect in persons.

This (part of) documentation was generated from src/fst/morphology/affixes/verbs.lexc

src-fst-morphology-phonology.twolc.md

=================================== ! The Lithuanian morphophonological/twolc rules file ! =================================== !

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

src-fst-morphology-root.lexc.md

INTRODUCTION TO MORPHOLOGICAL ANALYSER OF Lithuanian LANGUAGE.

Definitions for Multichar_Symbols

Analysis symbols

The morphological analyses of wordforms for the Lithuanian language are presented in this system in terms of the following symbols. (It is highly suggested to follow existing standards when adding new tags).

The parts-of-speech are:

The parts of speech are further split up into:

The Usage extents are marked using following tags:

The nominals are inflected in the following Case and Number

The possession is marked as such: The comparative forms are: Numerals are classified under: Verb moods are: Verb personal forms are: Other verb forms are

Question and Focus particles:

Semantics are classified with

Derivations are classified under the morphophonetic form of the suffix, the source and target part-of-speech.

Morphophonology To represent phonologic variations in word forms we use the following symbols in the lexicon files:

And following triggers to control variation

Flag diacritics

We have manually optimised the structure of our lexicon using following flag diacritics to restrict morhpological combinatorics - only allow compounds with verbs if the verb is further derived into a noun again: | @P.NeedNoun.ON@ | (Dis)allow compounds with verbs unless nominalised | @D.NeedNoun.ON@ | (Dis)allow compounds with verbs unless nominalised | @C.NeedNoun@ | (Dis)allow compounds with verbs unless nominalised

For languages that allow compounding, the following flag diacritics are needed to control position-based compounding restrictions for nominals. Their use is handled automatically if combined with +CmpN/xxx tags. If not used, they will do no harm. | @P.CmpFrst.FALSE@ | Require that words tagged as such only appear first | @D.CmpPref.TRUE@ | Block such words from entering ENDLEX | @P.CmpPref.FALSE@ | Block these words from making further compounds | @D.CmpLast.TRUE@ | Block such words from entering R | @D.CmpNone.TRUE@ | Combines with the next tag to prohibit compounding | @U.CmpNone.FALSE@ | Combines with the prev tag to prohibit compounding | @P.CmpOnly.TRUE@ | Sets a flag to indicate that the word has passed R | @D.CmpOnly.FALSE@ | Disallow words coming directly from root.

Use the following flag diacritics to control downcasing of derived proper nouns (e.g. Finnish Pariisi -> pariisilainen). See e.g. North Sámi for how to use these flags. There exists a ready-made regex that will do the actual down-casing given the proper use of these flags. | @U.Cap.Obl@ | Allowing downcasing of derived names: deatnulasj. | @U.Cap.Opt@ | Allowing downcasing of derived names: deatnulasj.

The word forms in Lithuanian language start from the lexeme roots of basic word classes, or optionally from prefixes:

This (part of) documentation was generated from src/fst/morphology/root.lexc

src-fst-morphology-stems-adjectives.lexc.md

Adjectives Adjectives in the Lithuanian language describe things.

This (part of) documentation was generated from src/fst/morphology/stems/adjectives.lexc

src-fst-morphology-stems-nouns.lexc.md

Nouns Nouns in the Lithuanian language are things.

This (part of) documentation was generated from src/fst/morphology/stems/nouns.lexc

src-fst-morphology-stems-numerals.lexc.md

Numerals Numerals in the Lithuanian language are numbers.

This (part of) documentation was generated from src/fst/morphology/stems/numerals.lexc

src-fst-morphology-stems-prefixes.lexc.md

Prefixes Prefixes in the Lithuanian language are bound to beginning of other words.

This (part of) documentation was generated from src/fst/morphology/stems/prefixes.lexc

src-fst-morphology-stems-pronouns.lexc.md

Pronouns Pronouns in the Lithuanian language are references to things.

This (part of) documentation was generated from src/fst/morphology/stems/pronouns.lexc

src-fst-morphology-stems-verbs.lexc.md

Verbs Verbs in the Lithuanian language are actions.

This (part of) documentation was generated from src/fst/morphology/stems/verbs.lexc

src-fst-phonetics-txt2ipa.xfscript.md

retroflex plosive, voiceless t ʈ 0288, 648 ( = ASCII 096) retroflex plosive, voiced d ɖ 0256, 598 labiodental nasal F ɱ 0271, 625 retroflex nasal n ɳ 0273, 627 palatal nasal J ɲ 0272, 626 velar nasal N ŋ 014B, 331 uvular nasal N\ ɴ 0274, 628

bilabial trill B\ ʙ 0299, 665 uvular trill R\ ʀ 0280, 640 alveolar tap 4 ɾ 027E, 638 retroflex flap r ɽ 027D, 637 bilabial fricative, voiceless p\ ɸ 0278, 632 bilabial fricative, voiced B β 03B2, 946 dental fricative, voiceless T θ 03B8, 952 dental fricative, voiced D ð 00F0, 240 postalveolar fricative, voiceless S ʃ 0283, 643 postalveolar fricative, voiced Z ʒ 0292, 658 retroflex fricative, voiceless s ʂ 0282, 642 retroflex fricative, voiced z` ʐ 0290, 656 palatal fricative, voiceless C ç 00E7, 231 palatal fricative, voiced j\ ʝ 029D, 669 velar fricative, voiced G ɣ 0263, 611 uvular fricative, voiceless X χ 03C7, 967 uvular fricative, voiced R ʁ 0281, 641 pharyngeal fricative, voiceless X\ ħ 0127, 295 pharyngeal fricative, voiced ?\ ʕ 0295, 661 glottal fricative, voiced h\ ɦ 0266, 614

alveolar lateral fricative, vl. K alveolar lateral fricative, vd. K\

labiodental approximant P (or v) alveolar approximant r\ retroflex approximant r` velar approximant M\

retroflex lateral approximant l` palatal lateral approximant L velar lateral approximant L
Clicks

bilabial O\ (O = capital letter) dental |
(post)alveolar !\ palatoalveolar =\ alveolar lateral ||
Ejectives, implosives

ejective > e.g. ejective p p> implosive < e.g. implosive b b< Vowels

close back unrounded M close central unrounded 1 close central rounded } lax i I lax y Y lax u U

close-mid front rounded 2 close-mid central unrounded @\ close-mid central rounded 8 close-mid back unrounded 7

schwa ə @

open-mid front unrounded E open-mid front rounded 9 open-mid central unrounded 3 open-mid central rounded 3\ open-mid back unrounded V open-mid back rounded O

ash (ae digraph) { open schwa (turned a) 6

open front rounded & open back unrounded A open back rounded Q Other symbols

voiceless labial-velar fricative W voiced labial-palatal approx. H voiceless epiglottal fricative H\ voiced epiglottal fricative <\ epiglottal plosive >\

alveolo-palatal fricative, vl. s\ alveolo-palatal fricative, voiced z\ alveolar lateral flap l\ simultaneous S and x x\ tie bar _ Suprasegmentals

primary stress “ secondary stress % long : half-long :\ extra-short _X linking mark -
Tones and word accents

level extra high _T level high _H level mid _M level low _L level extra low _B downstep ! upstep ^ (caret, circumflex)

contour, rising contour, falling _F contour, high rising _H_T contour, low rising _B_L

contour, rising-falling _R_F (NB Instead of being written as diacritics with _, all prosodic marks can alternatively be placed in a separate tier, set off by < >, as recommended for the next two symbols.) global rise global fall Diacritics

voiceless 0 (0 = figure), e.g. n_0 voiced _v aspirated _h more rounded _O (O = letter) less rounded _c advanced _+ retracted _- centralized _” syllabic = (or _=) e.g. n= (or n=) non-syllabic _^ rhoticity `

breathy voiced _t creaky voiced _k linguolabial _N labialized _w palatalized ‘ (or _j) e.g. t’ (or t_j) velarized _G pharyngealized _?\

dental d apical _a laminal _m nasalized ~ (or _~) e.g. A~ (or A~) nasal release _n lateral release _l no audible release _}

velarized or pharyngealized _e velarized l, alternatively 5 raised _r lowered _o advanced tongue root _A retracted tongue root _q

This (part of) documentation was generated from src/fst/phonetics/txt2ipa.xfscript

src-fst-transcriptions-transcriptor-abbrevs2text.lexc.md

We describe here how abbreviations are in Lithuanian are read out, e.g. for text-to-speech systems.

For example:

This (part of) documentation was generated from src/fst/transcriptions/transcriptor-abbrevs2text.lexc

src-fst-transcriptions-transcriptor-numbers-digit2text.lexc.md

% komma% :, Root ; % tjuohkkis% :%. Root ; % kolon% :%: Root ; % sárggis% :%- Root ; % násti% :%* Root ;

This (part of) documentation was generated from src/fst/transcriptions/transcriptor-numbers-digit2text.lexc

tools-tokenisers-tokeniser-disamb-gt-desc.pmscript.md

Tokeniser for lit

Usage:

$ make
$ echo "ja, ja" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
$ echo "Juos gorreválggain lea (dárbbašlaš) deavdit gáibádusa boasttu olmmoš, man mielde lahtuid." | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
$ echo "(gáfe) 'ja' ja 3. ja? ц jaja ukjend \"ukjend\"" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
$ echo "márffibiillagáffe" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst

Pmatch documentation: https://github.com/hfst/hfst/wiki/HfstPmatch

Characters which have analyses in the lexicon, but can appear without spaces before/after, that is, with no context conditions, and adjacent to words:

Whitespace contains ASCII white space and the List contains some unicode white space characters

Apart from what’s in our morphology, there are

  1. unknown word-like forms, and
  2. unmatched strings We want to give 1) a match, but let 2) be treated specially by hfst-tokenise -a Unknowns are made of:
    • lower-case ASCII
    • upper-case ASCII
    • select extended latin symbols ASCII digits
    • select symbols
    • Combining diacritics as individual symbols,
    • various symbols from Private area (probably Microsoft), so far:
    • U+F0B7 for “x in box”

Unknown handling

Unknowns are tagged ?? and treated specially with hfst-tokenise hfst-tokenise –giella-cg will treat such empty analyses as unknowns, and remove empty analyses from other readings. Empty readings are also legal in CG, they get a default baseform equal to the wordform, but no tag to check, so it’s safer to let hfst-tokenise handle them.

Finally we mark as a token any sequence making up a:

This (part of) documentation was generated from tools/tokenisers/tokeniser-disamb-gt-desc.pmscript

tools-tokenisers-tokeniser-gramcheck-gt-desc.pmscript.md

Grammar checker tokenisation for lit

Requires a recent version of HFST (3.10.0 / git revision>=3aecdbc) Then just:

$ make
$ echo "ja, ja" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst

More usage examples:

$ echo "Juos gorreválggain lea (dárbbašlaš) deavdit gáibádusa boasttu olmmoš, man mielde lahtuid." | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
$ echo "(gáfe) 'ja' ja 3. ja? ц jaja ukjend \"ukjend\"" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
$ echo "márffibiillagáffe" | hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst

Pmatch documentation: https://github.com/hfst/hfst/wiki/HfstPmatch

Characters which have analyses in the lexicon, but can appear without spaces before/after, that is, with no context conditions, and adjacent to words:

Whitespace contains ASCII white space and the List contains some unicode white space characters

Apart from what’s in our morphology, there are 1) unknown word-like forms, and 2) unmatched strings We want to give 1) a match, but let 2) be treated specially by hfst-tokenise -a

TODO: Could use something like this, but built-in’s don’t include šžđčŋ:

Simply give an empty reading when something is unknown: hfst-tokenise –giella-cg will treat such empty analyses as unknowns, and remove empty analyses from other readings. Empty readings are also legal in CG, they get a default baseform equal to the wordform, but no tag to check, so it’s safer to let hfst-tokenise handle them.

Finally we mark as a token any sequence making up a:

This (part of) documentation was generated from tools/tokenisers/tokeniser-gramcheck-gt-desc.pmscript

tools-tokenisers-tokeniser-tts-cggt-desc.pmscript.md

TTS tokenisation for smj

Requires a recent version of HFST (3.10.0 / git revision>=3aecdbc) Then just:

make
echo "ja, ja" \
| hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst

More usage examples:

echo "Juos gorreválggain lea (dárbbašlaš) deavdit gáibádusa \
boasttu olmmoš, man mielde lahtuid." \
| hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
echo "(gáfe) 'ja' ja 3. ja? ц jaja ukjend \"ukjend\"" \
| hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst
echo "márffibiillagáffe" \
| hfst-tokenise --giella-cg tokeniser-disamb-gt-desc.pmhfst

Pmatch documentation: https://kitwiki.csc.fi/twiki/bin/view/KitWiki/HfstPmatch

Characters which have analyses in the lexicon, but can appear without spaces before/after, that is, with no context conditions, and adjacent to words:

Whitespace contains ASCII white space and the List contains some unicode white space characters

Apart from what’s in our morphology, there are 1) unknown word-like forms, and 2) unmatched strings We want to give 1) a match, but let 2) be treated specially by hfst-tokenise -a

TODO: Could use something like this, but built-in’s don’t include šžđčŋ:

Simply give an empty reading when something is unknown: hfst-tokenise –giella-cg will treat such empty analyses as unknowns, and remove empty analyses from other readings. Empty readings are also legal in CG, they get a default baseform equal to the wordform, but no tag to check, so it’s safer to let hfst-tokenise handle them.

Needs hfst-tokenise to output things differently depending on the tag they get

This (part of) documentation was generated from tools/tokenisers/tokeniser-tts-cggt-desc.pmscript