Mova0.6 pre-release

the flag of India 8. Word classes and Universal Dependencies tags

Kannada word classes are identified by the forms a word can take and by the positions and constructions in which it occurs. Universal Dependencies (UD) records the class of each syntactic word as a universal part-of-speech tag, or UPOS, and records properties such as case and number as features. A word’s class, its features and its relation to another word in the sentence are three distinct parts of the analysis.

Distribution and morphology as word-class diagnostics

Kannada’s usual order in a clause with an overt subject and object is subject–object–verb. This order helps identify the predicate in an ordinary clause, but it is a poor test of the class of a word in isolation. Several classes can occur before nouns, and a clause may have a nominal predicate. Kannada also has substantial inflection, so suffixes, agreement and the distribution of complete forms provide evidence alongside word order. (Wikibooks contributors 2026, § Grammar.)

DiagnosticEvidence it providesCondition on its use
Case and number inflectionIdentifies a nominal use, including common nouns, names and pronounsA singular or nominative form may lack an overt ending. A nominal use can also arise through derivation from another class.
Tense and agreementIdentifies an inflecting verbal predicate or the inflecting member of a verbal complexNon-finite verbs have a different set of endings. Imperatives also have their own forms.
Position before a nounIdentifies a candidate for a noun modifierAdjectives, determiners, numerals, genitives and verbal modifiers all occur in this position.
Modification of a predicate or clauseIdentifies a candidate for adverbial functionAn adverb, a case-marked noun, a phrase with a postposition or a non-finite clause can have this function.
Ability to fill an argument position aloneHelps identify nouns and pronouns and shows when another form is being used independentlyA numeral or a demonstrative can also occur without an overt noun. Examine the form and its construction before assigning a class.
Scope over a word, phrase or clauseHelps identify a particle or a connectiveThe same item may have several uses. Determine what it modifies or links in the particular sentence.

Case illustrates the difference between class and function. In the first example below, the accusative noun is the object of ‘see’. In the second, a locative noun gives the setting of ‘study’. Both inflected words remain nouns. In the third, a noun itself supplies the predicate.

FormRomanizationGlossTranslation
ಅವಳು ಮನೆಯನ್ನು ನೋಡಿದಳು.avaḷu maneyannu nōḍidaḷushe house-ACC see-PST.3SG.F‘She saw the house.’ (constructed)
ಅವಳು ಮನೆಯಲ್ಲಿ ಓದಿದಳು.avaḷu maneyalli ōdidaḷushe house-LOC study-PST.3SG.F‘She studied at home.’ (constructed)
ಅವಳು ಶಿಕ್ಷಕಿ.avaḷu śikṣakishe teacher.F‘She is a teacher.’ (constructed)

Read the glosses from left to right in word order. A hyphen shows a grammatical element within a word, and periods group several properties of one form. ACC is accusative, DAT dative, GEN genitive and LOC locative. SG and PL are singular and plural, M, F and N are masculine, feminine and neuter, and 1, 2 and 3 identify person. PST is past, NPST nonpast, IMP imperative, INF infinitive, CVB converb and REL a verbal form that modifies a noun. QUOT is a quotative marker and FOC a focus marker. These glosses describe examples. The UD feature names and values appear in section 6.

Nouns, proper nouns, pronouns and determiners

Common nouns and proper names

A common noun names a kind, substance, person, place or other entity by category. A proper noun identifies an individual, a place or another named entity by name. The UD distinction is NOUN versus PROPN. Kannada spelling has no uppercase contrast to mark names. Names can take case endings just as common nouns do, so both the meaning of the expression and its use in context matter for the distinction. The Kannada Wikibooks noun section illustrates nouns with personal and geographical names. (Wikibooks contributors 2026, § Noun.)

UseFormRomanizationGlossTranslation
Common noun, singularಮನೆmanehouse‘house’ (constructed)
Common noun, pluralಮನೆಗಳುmanegaḷuhouse-PL‘houses’ (constructed)
Common noun, singular dativeಮನೆಗೆmanegehouse-DAT‘to the house’ (constructed)
Proper name used as a destinationಬೆಂಗಳೂರಿಗೆbeṅgaḷūrigeBengaluru-DAT‘to Bengaluru’ (constructed)

The number contrast is singular versus plural. The modern literary gender classification described by Wikibooks distinguishes masculine, feminine and neuter. In its account, nouns denoting men and boys are masculine, nouns denoting women and girls are feminine, and remaining nouns fall under neuter. Agreement provides a separate observation about a particular construction. For example, a predicate may show feminine singular or neuter singular agreement even when the subject noun has no overt gender suffix. Treat the traditional classification of nouns and the agreement features visible in a sentence as separate evidence. (Wikibooks contributors 2026, §§ Grammar, Gender.)

FormRomanizationGlossTranslation
ಹುಡುಗ ಬಂದನು.huḍuga bandanuboy come-PST.3SG.M‘The boy came.’ (constructed)
ಹುಡುಗಿ ಬಂದಳು.huḍugi bandaḷugirl come-PST.3SG.F‘The girl came.’ (constructed)
ನಾಯಿ ಬಂತು.nāyi bantudog come-PST.3SG.N‘The dog came.’ (constructed)

Pronouns and determiners

Personal pronouns such as ‘I’ or ‘she’ can fill an argument position without a noun. They belong to PRON. Their inflected forms can have a different stem from the subject form. A genitive pronoun can serve as a possessor within a noun phrase, as in ‘my house’. Its position before the noun is compatible with that analysis. Kannada nominal inflection and the pronoun system are treated separately in Kittel’s overview of the grammar. (Kittel 1903, §§ IV, XIII.)

First-person singular useFormRomanizationGlossTranslation
Subjectನಾನುnānu1SG.NOM‘I’ (constructed)
Accusativeನನ್ನನ್ನುnannannu1SG.ACC‘me’ (constructed)
Dativeನನಗೆnanage1SG.DAT‘to me’ (constructed)
Genitiveನನ್ನnanna1SG.GEN‘my’ (constructed)

A determiner such as the adnominal demonstrative ‘this’ identifies which entity the noun phrase refers to. Use DET in that construction. An independent demonstrative like ‘this one’ can be a subject or object and is a pronoun. The near and far distinctions in the following examples are visible both in adnominal and in independent forms.

UseFormRomanizationGlossTranslation
Near, before a nounಈ ಪುಸ್ತಕī pustakathis book‘this book’ (constructed)
Far, before a nounಆ ಪುಸ್ತಕā pustakathat book‘that book’ (constructed)
Near, independent, neuter singularಇದುiduthis.N.SG‘this one’ (constructed)
Far, independent, neuter singularಅದುaduthat.N.SG‘that one’ (constructed)
Genitive possessorನನ್ನ ಪುಸ್ತಕnanna pustaka1SG.GEN book‘my book’ (constructed)
Interrogative determinerಯಾವ ಪುಸ್ತಕyāva pustakawhich book‘which book’ (constructed)
Interrogative pronounಯಾರುyāruwho‘who’ (constructed)

A phrase headed by a bare noun can receive a definite or an indefinite interpretation in context. The English articles used in translations supply the interpretation intended for the example. They do not supply a Kannada word to tag. A cardinal numeral that counts entities is NUM, including when English could translate ‘one’ as an indefinite article. In uses where a form has specifically acquired a determiner function, identify that function from the construction. Kittel lists the expression of article meanings separately from numerals. (Kittel 1903, §§ XV, XVIII.)

FormRomanizationGlossTranslation
ಒಂದು ಪುಸ್ತಕondu pustakaone book‘one book’ (constructed)
ಎರಡು ಪುಸ್ತಕಗಳುeraḍu pustakagaḷutwo book-PL‘two books’ (constructed)

Verbs, auxiliaries and non-finite forms

Finite verbs

A lexical verb supplies the predicate’s event or state and contributes to the choice of arguments. In an independent clause, a finite verb can carry tense and, in forms that distinguish them, person, number and gender. An imperative can serve as an independent predicate with an understood addressee. The citation forms in a verb list can coincide with familiar singular imperatives. Their translation by English ‘to’ plus a verb gives the lexical meaning and does not identify their Kannada morphology. Kannada Wikibooks gives forms for basic verbs such as ‘come’, ‘go’, ‘read’ and ‘do’. (Kittel 1903, §§ V, XIX; Wikibooks contributors 2026, § Verb.)

UseFormRomanizationGlossTranslation
Past, first-person singularನಾನು ಓದಿದೆ.nānu ōdide1SG read-PST.1SG‘I read it.’ (constructed)
Past, third-person feminine singularಅವಳು ಓದಿದಳು.avaḷu ōdidaḷushe read-PST.3SG.F‘She read it.’ (constructed)
Nonpast, third-person feminine singularಅವಳು ಓದುತ್ತಾಳೆ.avaḷu ōduttāḷeshe read-NPST.3SG.F‘She reads.’ (constructed)
Familiar singular imperativeಪುಸ್ತಕ ಓದು.pustaka ōdubook read.IMP‘Read the book.’ (constructed)

The overt subject and object can vary across these examples without changing the word class of ‘read’. The nonpast example has a habitual reading. A tense value describes the grammatical form under the adopted analysis. A habitual reading, a future time reference and an instruction can each involve additional information from the construction and context.

Auxiliaries and verbal complexes

Use VERB for a verb in a lexical predicate use. AUX is available for a syntactic word with a grammatical role in a verbal complex, such as marking aspect or modality, and for an overt copula where the analysis treats it as such. A form with an independent lexical use can also serve as an auxiliary. A second verb may instead introduce a second event or a verbal complement. Inflection on the second form alone therefore does not decide between analyses. Kittel treats auxiliary uses and combinations of verbs in separate parts of his grammar. (Kittel 1903, §§ XX, XXIV.)

ConstructionFormRomanizationGlossTranslation
Independent locative predicateಅವಳು ಮನೆಯಲ್ಲಿದ್ದಾಳೆ.avaḷu maneyalliddāḷeshe house-LOC-be.3SG.F‘She is at home.’ (constructed)
Progressive with the same verb stem for ‘be’ಅವಳು ಓದುತ್ತಿದ್ದಾಳೆ.avaḷu ōduttiddāḷeshe read-IPFV-be.3SG.F‘She is reading.’ (constructed)
Two events in sequenceಅವಳು ಊಟ ಮಾಡಿ ಮಲಗಿದಳು.avaḷu ūṭa māḍi malagidaḷushe meal do.CVB go.to.bed-PST.3SG.F‘She ate and went to bed.’ (constructed)

IPFV identifies the imperfective element in the gloss. The first sentence locates the subject. In the second, the element meaning ‘be’ contributes to the progressive of ‘read’. In the third, eating and going to bed are two events in sequence. The examples also illustrate why segmentation matters. An analysis of the progressive as containing an auxiliary element does not by itself decide whether that element receives its own UPOS tag. UD assigns tags after the syntactic words have been identified.

Non-finite verbs

Non-finite forms can retain the arguments and modifiers of a lexical verb while serving in a larger construction. An infinitive can head a complement or express purpose. A converb can link an event to a following predicate, with the precise temporal or other relation supplied by form and context. A verbal relative form can modify a noun and take its own object or adverbial. In these uses the lexical verb has the tag VERB. Its non-finite form and its role in the larger construction are recorded separately.

ConstructionFormRomanizationGlossTranslation
Infinitive expressing purposeಅವಳು ಪುಸ್ತಕ ಓದಲು ಬಂದಳು.avaḷu pustaka ōdalu bandaḷushe book read-INF come-PST.3SG.F‘She came to read a book.’ (constructed)
Converb preceding the main eventಅವಳು ಪುಸ್ತಕ ಓದಿ ಮಲಗಿದಳು.avaḷu pustaka ōdi malagidaḷushe book read.CVB go.to.bed-PST.3SG.F‘She read the book and went to bed.’ (constructed)
Verbal relative with an overt objectಪುಸ್ತಕ ಓದಿದ ಹುಡುಗಿpustaka ōdida huḍugibook read-PST.REL girl‘the girl who read the book’ (constructed)
Verbal relative with an understood objectಅವಳು ಓದಿದ ಪುಸ್ತಕavaḷu ōdida pustakashe read-PST.REL book‘the book she read’ (constructed)

A derived noun can denote an activity as well. Diagnostics for the noun include case inflection and the distribution of a nominal head. Where the form heads a verbal construction with its own arguments, inspect that construction before deciding whether case is borne by a nominalized verb or by a lexical noun. Kittel’s outline distinguishes verbal forms that take nominal inflection from the main discussion of verbs. (Kittel 1903, §§ V, XII.)

Adjectives, adverbs and attributive forms

Adjectives and nominal modifiers

A lexical adjective describes a property of a noun’s referent. Ordinary prenominal adjectives can keep the same form while the noun changes for number or case. The table follows one adjective across three noun forms.

Noun phraseFormRomanizationGlossTranslation
Singularದೊಡ್ಡ ಮನೆdoḍḍa manebig house‘a big house’ (constructed)
Pluralದೊಡ್ಡ ಮನೆಗಳುdoḍḍa manegaḷubig house-PL‘big houses’ (constructed)
Singular dativeದೊಡ್ಡ ಮನೆಗೆdoḍḍa manegebig house-DAT‘to a big house’ (constructed)

Use ADJ for this kind of lexical property modifier. Other words before a noun can contribute a similar description but show different morphology or internal structure. An inflected genitive retains the class of its nominal base. A relative verb can have the noun as its understood subject, object or another participant in the event. The dependency of a complete expression on the noun consequently does not establish ADJ for the word heading the expression. Kittel discusses adjectives, verbal declinable forms and comparison in separate sections. (Kittel 1903, §§ XII, XIV, XXV.)

Source of the modifierFormRomanizationGlossTranslation
Lexical adjectiveಹಳೆಯ ಪುಸ್ತಕhaḷeya pustakaold book‘an old book’ (constructed)
Genitive nounಶಾಲೆಯ ಪುಸ್ತಕśāleya pustakaschool-GEN book‘the school’s book’ (constructed)
Verbal relative with a time modifierನಿನ್ನೆ ಬಂದ ಹುಡುಗninne banda huḍugayesterday come.PST.REL boy‘the boy who came yesterday’ (constructed)

Adverbs and adverbial expressions

A lexical adverb can express time, manner, degree or another circumstance. Use ADV when the word itself belongs to this class. Similar meanings can be expressed by case-marked nominals, phrases with postpositions and non-finite clauses. In those expressions the syntactic head retains its own word class. Some adverbs can modify an adjective or another adverb as well as a verb. (Kittel 1903, §§ VI, XVI.)

UseFormRomanizationGlossTranslation
Lexical time adverbಅವಳು ನಿನ್ನೆ ಬಂದಳು.avaḷu ninne bandaḷushe yesterday come-PST.3SG.F‘She came yesterday.’ (constructed)
Manner adverbಅವಳು ಬೇಗ ಬಂದಳು.avaḷu bēga bandaḷushe quickly come-PST.3SG.F‘She came quickly.’ (constructed)
Degree adverb modifying an adjectiveಬಹಳ ದೊಡ್ಡ ಮನೆbahaḷa doḍḍa manevery big house‘a very big house’ (constructed)
Locative noun used adverbiallyಅವಳು ಮನೆಯಲ್ಲಿ ಮಲಗಿದಳು.avaḷu maneyalli malagidaḷushe house-LOC sleep-PST.3SG.F‘She slept at home.’ (constructed)

A lexical degree adverb modifies gradable content in the last example with a noun phrase. Its presence does not require a degree feature on the adjective. Assign an inflectional feature only where the form bears the relevant grammatical distinction.

Adpositions, conjunctions, particles and interjections

These classes have different domains. An adposition introduces a relation involving a nominal expression. A coordinator links units of equal syntactic status, while a subordinator marks a clause as dependent. A particle contributes a grammatical or discourse meaning within its scope. An interjection can stand as a reaction or call without a clause.

Adpositions

Kannada adpositions in the constructions illustrated here follow their complements. The complement can show a case selected by the postposition. In the examples below it has a genitive form with ‘about’ and ‘after’. Tag a separate word in postpositional use ADP and record the case on the nominal where the inflection belongs. A case suffix is part of the inflected word. For a form that also occurs independently as a noun or adverb, identify whether the construction uses it to mark a relation to an overt complement. (Kittel 1903, §§ IV, VI, XVI.)

FormRomanizationGlossTranslation
ಪುಸ್ತಕದ ಬಗ್ಗೆpustakada baggebook-GEN about‘about the book’ (constructed)
ಊಟದ ನಂತರūṭada nantarameal-GEN after‘after the meal’ (constructed)

Coordinators, subordinators and clause linkage

Use CCONJ for a separate word that marks coordination of comparable units, including words, phrases or clauses. Use SCONJ for a separate word whose role is to mark a dependent clause. Kannada can also express relations between clauses through verbal morphology, as with the converb in section 3. The verb in such a form retains its verbal tag. A translation with English ‘and’, ‘after’ or ‘if’ is insufficient evidence of a conjunction word in Kannada. Kittel’s contents treat paired connectives, quotative verbs and combinations of verbal forms under separate headings. (Kittel 1903, §§ XXI, XXII, XXIV.)

ConstructionFormRomanizationGlossTranslation
Coordination of nounsಹಾಲು ಮತ್ತು ನೀರುhālu mattu nīrumilk and water‘milk and water’ (constructed)
Alternative coordinationಹಾಲು ಅಥವಾ ನೀರುhālu athavā nīrumilk or water‘milk or water’ (constructed)
Conditional verbal morphologyಮಳೆ ಬಂದರೆ ನಾವು ಹೋಗುವುದಿಲ್ಲ.maḷe bandare nāvu hōguvudillarain come-COND 1PL go-NONFIN.NEG‘If it rains, we will not go.’ (constructed)
Overt quotative marker after a clauseಅವಳು ಬರುತ್ತಾಳೆ ಎಂದು ಅವನು ಹೇಳಿದನು.avaḷu baruttāḷe endu avanu hēḷidanushe come-NPST.3SG.F QUOT he say-PST.3SG.M‘He said that she would come.’ (constructed)

COND labels a conditional form, NONFIN a non-finite element and NEG negation. In the last example, ‘say’ is the reporting predicate and the quotative marks the reported clause. When a separately segmented quotative is analyzed as a dependent-clause marker, SCONJ expresses that role. Related forms used as predicates of saying have verbal distribution.

Particles and interjections

PART covers separately tagged grammatical items such as a focus particle in a use that singles out a constituent. Focus and emphasis can also be expressed by bound elements. When a focus marker is written with its host, first determine whether the segmentation treats it as its own syntactic word. If the host and marker form one syntactic word, that word takes the host’s class. If the marker is a separate syntactic word, its scope can extend over the host or the larger phrase containing it. Kittel treats expressions of exclusivity and interjections in separate sections. (Kittel 1903, §§ XVII, XVIII.)

UseFormRomanizationGlossTranslation
Additive focusಅವಳೂ ಬಂದಳು.avaḷū bandaḷushe-ADD come-PST.3SG.F‘She came too.’ (constructed)
Focus on a personal pronounಅವಳೇ ಬಂದಳು.avaḷē bandaḷushe-FOC come-PST.3SG.F‘She herself came.’ (constructed)
Independent reactionಅಯ್ಯೋ!ayyōalas‘Oh no!’ (constructed)

ADD labels additive meaning. The host in each of the first two examples is a personal pronoun. The third example has the tag INTJ in its use as an independent reaction. An exclamation mark is PUNCT if represented as a separate token. Its presence after ordinary declarative or imperative words leaves their word classes intact.

Universal Dependencies tags and features

UPOS supplies one class for each syntactic word. Use the distribution of the particular occurrence in context. A local part-of-speech label can be recorded separately in XPOS. FEATS records grammatical properties of the analyzed word, and the dependency relation records its role in the tree. The following table gives the complete UPOS inventory. The mappings describe the tag categories and how the Kannada diagnostics above apply.

UPOSClassUse
ADJAdjectiveA lexical property word, including a lexical adjectival modifier of a noun.
ADPAdpositionA relational word with a nominal complement, typically following it in the Kannada constructions illustrated here.
ADVAdverbA lexical modifier expressing a circumstance, degree or related meaning.
AUXAuxiliaryA separately represented auxiliary in a verbal complex or a word analyzed as a copula.
CCONJCoordinating conjunctionA word marking coordination of comparable units.
DETDeterminerAn adnominal determiner, such as a demonstrative that identifies the noun’s referent.
INTJInterjectionA conventional reaction, call or other interjectional expression.
NOUNCommon nounA nominal head such as a word for a kind of entity or a lexical noun derived from a verb.
NUMNumeralA cardinal used as a number or to count entities, whether written in words or digits. An ordinal with adjectival distribution takes ADJ.
PARTParticleA syntactic word with a particle function, such as focus marking in the construction at hand.
PRONPronounA personal pronoun, an independent demonstrative or another pronominal word.
PROPNProper nounA name or a word with proper-name status within a name. Case inflection is compatible with this tag.
PUNCTPunctuationA token serving as punctuation.
SCONJSubordinating conjunctionA word whose function is to mark a dependent clause.
SYMSymbolA token functioning as a symbol, such as a currency sign. A digit string functioning as a cardinal takes NUM.
VERBVerbA lexical verb, including one in an imperative, an infinitive, a converb or a verbal relative construction.
XOtherA token that cannot be assigned one of the preceding classes, for example genuinely unclassifiable material. A loanword with ordinary Kannada noun distribution can be assigned NOUN.

Features and the forms that bear them

Use a feature where the word’s analysis supports it. A narrowly defined morphological value can be justified even when its surface realization is zero. Where the form is ambiguous, use context and agreement when they resolve the analysis. The table lists features relevant to this chapter and gives representative values for the contrasts discussed here.

FeatureValues relevant to the examplesWhere to look for evidence
CaseNom, Acc, Dat, Gen, LocNominal inflection and the construction. For example, the accusative and locative forms of ‘house’ in section 1 bear different values.
NumberSing, PlurThe nominal number contrast, the pronoun paradigm or the number realized by verbal agreement.
GenderMasc, Fem, NeutA gender distinction supported by the nominal or pronominal analysis, or by the agreeing form itself.
Person1, 2, 3A personal pronoun or a verb form with person agreement. A noun can control third-person agreement without receiving a Person feature itself.
PronTypePrs, Dem, IntPersonal, demonstrative and interrogative uses. A demonstrative can carry PronType=Dem with either DET or PRON.
PossYesA possessive form under an analysis that encodes possession with this feature. Genitive case and possessive function are distinct properties.
NumTypeCard, OrdA cardinal or an ordinal reading under the analysis of the numeral expression.
VerbFormFin, Inf, Conv, PartThe finite or non-finite paradigm. Conv can represent a converb and Part a verbal relative form analyzed as a participle.
MoodImpAn imperative form. A command expressed through another form still has the morphology of that form.
TensePast, Pres, FutA grammatical tense distinction under the adopted analysis. The time given by an adverb or inferred from the discourse alone is insufficient.
AspectProgA grammatical progressive where the annotation represents aspect on a word. The organization of an auxiliary construction depends on its segmentation.
PolarityNegThe analyzed word that expresses grammatical negation. Scope over a phrase does not call for copying the feature to every word in it.

For instance, in the first example of the chapter, the object ‘house’ is NOUN with Case=Acc and Number=Sing. Its role as object can be represented by obj. The feminine singular past verb is VERB and supports Gender=Fem, Number=Sing, Person=3, Tense=Past and VerbForm=Fin. In the locative example, ‘house’ remains NOUN. It has Case=Loc and serves as an oblique, which can be represented by obl. The tag thus remains stable across two different inflections and dependency roles.

The same reasoning applies within a noun phrase. An adjective can be an amod, an adnominal demonstrative a det, a cardinal modifier a nummod, and a verbal relative clause an acl. A postposition can be an ADP with the dependency relation case, while its complement carries the morphological feature Case when the nominal form supports it. PART, the word-class tag for a particle, and VerbForm=Part, the feature for a participle, describe different things.

How Mova handles Kannada word classes

Kannada (kn) has no level-2 parser model. The first level of tools/mova-parse uses tools/langseg to assign languages to sentences and spans. Kannada script falls under other. The second level uses a language-specific model to parse a span. For Kannada the parser reports the missing model in the # parse comment and leaves _ in LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL, with Parse=None in MISC. Mova does not tokenize, tag, lemmatize or parse Kannada. It holds no annotated Kannada corpus and has no measured Kannada treebank baseline.

The baseline requested in What every language gets uses the en crate. en train trains a tree parser and tagger on a language’s UD training split after en xpos-fill. en parse-eval measures unlabeled and labeled attachment scores on the development split. prag and world read English trees and have no Kannada parse to read.

Squid reads the grammar and turns grammatical points into claims. Formalizable claims wait in Squid’s queue. The word-class and feature rules in this chapter have no implementation for Kannada. The first points for implementation are those needed by a tokenizer, a tagger and a UD treebank baseline.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/kn/08-word-classes/