Mova0.6 pre-release

the flag of India 22. Complementation and quotation

An Old Kannada predicate can take reported words, a proposition, or an event as its complement. A report can contain a finite clause, an action can be expressed by an infinitive, and a nominalized clause can take case in the larger sentence. The form of the dependent predicate, the meaning of the governing predicate, and the source of the dependent subject together distinguish these constructions from purpose clauses and sequences of actions.

Finite and nominalized complements

A complement supplies content associated with a governing predicate. With a speech predicate it can supply what was said, and with a thought predicate it can supply what was thought. With a predicate of perception or knowledge, it can identify a perceived event or a fact. Such content can have participants of its own. The speaker, thinker, or perceiver belongs to the governing clause, while the subject of the reported or perceived event belongs to the content clause. The two subjects may coincide. Their being different is also compatible with complementation (Kittel 1903, § on the construction of verbs).

A finite predicate inside a report retains the inflection appropriate to its own clause. In the first example below, the arrival predicate has a third-person masculine singular subject, the messenger. The same person, number, and gender values occur on the governing predicate, whose subject is the king. The identical endings occur with two different subjects. The particle at the right edge of the reported content is the quotative form eṃdu, historically a participial form of the verb ‘say’. The report as a whole depends on the governing predicate (Kittel 1903, § on the verbal participles; Kittel 1894, § ಎನ್ನು).

ConstructionFormTransliterationGlossTranslation
Finite content with a quotativeದೂತಂ ಬಂದಂ ಎಂದು ಅರಸಂ ನೆನೆದಂdūtaṃ baṃdaṃ eṃdu arasaṃ nenedaṃmessenger.NOM come.PST.3SG.M QUOT king.NOM think.PST.3SG.M‘The king thought that the messenger had come.’ (constructed)
Event nominalization in the accusativeದೂತಂ ಬಂದುದಂ ಅರಸಂ ಕಂಡಂdūtaṃ baṃdudaṃ arasaṃ kaṃḍaṃmessenger.NOM come.PST.NMLZ.ACC king.NOM see.PST.3SG.M‘The king saw the messenger’s arrival.’ (constructed)

Here and below, QUOT marks the quotative use, NMLZ a nominalization, and PST past. Person, number, and gender are glossed together as 3SG.M or 1SG where applicable. NOM, ACC, and DAT identify nominative, accusative, and dative.

A nominalized clause has nominal distribution externally and can retain verbal structure internally. The messenger is the participant who comes, and the past form presents the arrival as an event. The accusative on baṃdudaṃ belongs to the expression denoting the arrival. It gives that expression its place as the object of ‘see’. It does not assign accusative case to the messenger. Thus the relation of a participant to the dependent verb and the relation of the entire nominalization to the governing verb are determined at different levels (Kittel 1903, § on verbal nouns and the accusative).

Forms built with a neuter nominal element admit more than one interpretation. Depending on the lexical predicate and context, a form can refer to an occurrence, to the fact that something happened, or to an entity identified by what it did or what happened to it. A temporal modifier of the dependent event, a dependent object, or an expressed subject can help establish that the form has internal clause structure. The selecting predicate helps determine its interpretation. Perception can concern an occurrence, while knowledge can concern a fact. A translation that uses English that leaves this difference in the Kannada form open (Kittel 1903, § on verbal nouns and pronominal forms of participles).

An adnominal participle supplies a further structural comparison. In the following pair, the first expression can denote an arrival and take the object case of a governing predicate. In the second, the participle modifies an overt noun denoting the individual who arrived. The case of this noun phrase is the case of its head noun.

ConstructionFormTransliterationGlossTranslation
Nominalizationದೂತಂ ಬಂದುದುdūtaṃ baṃdudumessenger.NOM come.PST.NMLZ‘the messenger’s arrival’ (constructed)
Adnominal participle with a head nounಬಂದ ದೂತಂbaṃda dūtaṃcome.PST.ADN messenger.NOM‘the messenger who came’ (constructed)

The quotative example establishes a finite predicate within dependent content. For a finite string without a quotative or another overt indication of dependence, test whether the governing predicate selects the content and whether the surrounding passage treats the two as a single construction. Adjacent finite predicates may form successive assertions, and a speech passage can present an utterance alongside a reporting clause. An English translation with that gives the content relation intended by the translator. The Kannada wording and context supply the evidence for its structure.

Infinitival complements and understood subjects

The infinitive can express an action selected by a governing predicate or the purpose of a governing action. Its form supplies no person, number, or gender agreement for the dependent subject. An infinitive can still have the objects and other dependents licensed by its lexical verb. The action’s participant is recovered from an expressed argument, the lexical construction, or the discourse. The examples use the form in -al (Kittel 1903, § on the infinitive).

ConstructionFormTransliterationGlossTranslation
Action selected by a predicate of beginningಬರೆಯಲ್ ತೊಡಗಿದಂbareyal toḍagidaṃwrite.INF begin.PST.3SG.M‘He began to write.’ (constructed)
Action directed at an addresseeಅರಸಂ ದೂತಂಗೆ ಬರಲ್ ಪೇಳ್ದಂarasaṃ dūtaṃge baral pēḷdaṃking.NOM messenger.DAT come.INF tell.PST.3SG.M‘The king told the messenger to come.’ (constructed)
Purpose of motionನೋಡಲ್ ಬಂದಂnōḍal baṃdaṃlook.INF come.PST.3SG.M‘He came to look.’ (constructed)

In the first example, ‘begin’ locates the onset of the writing. The subject understood for writing is also the participant associated with the beginning. With the directive reading of the second example, the king gives the instruction and the messenger is expected to come. The dative identifies the addressee of the governing speech predicate. The directive interpretation connects that addressee to the action expressed by the infinitive. Dative case by itself can also identify a recipient of speech whose activity is independent of the reported content. In the third example, the motion predicate describes a coming, and the infinitive gives the purpose of that coming. The same participant is understood for both actions in this example, a pattern that also occurs with selected infinitives (Kittel 1903, § on infinitival constructions; Kittel 1894, § ತೊಡಗು and § ಪೇಳು).

Selection and purpose can be distinguished by asking what the dependent action contributes. In the beginning construction, writing is the activity whose onset is asserted. In the purpose construction, looking is the intended outcome of a motion event that the sentence already identifies. A purpose can be unrealized even when the motion has occurred. An order likewise need not be carried out, so lack of an assertion that the dependent action occurred cannot distinguish an infinitival complement from purpose. The decisive evidence comes from the relation of the action to the governing predicate in context.

For an omitted participant, start with the roles of the two lexical predicates. A speech predicate can have a speaker and an addressee, while ‘come’ has a participant who moves. Identify whom the passage presents as expected to move. In a purpose construction, the mover often supplies the actor of the purpose, but that reading follows from the relation between the events and the available participants. A discourse antecedent can be especially important when the governing subject or addressee is itself unexpressed. A single example with one salient individual leaves several explanations for coreference available.

A nominalized action can occupy a case-marked nominal position. An infinitive supplies an action without that nominal case marking. Both may be translated by an English infinitive or by an English expression in -ing. For comparing the Kannada constructions, record the form of the dependent predicate first, then its function and understood subject. This order also prevents the tense of the governing verb from being mistaken for an independently inflected tense of the infinitive. The temporal interpretation of writing in ‘began to write’ and of coming in ‘told him to come’ follows from the governing predicate and the construction.

Quotative constructions with speech and thought predicates

The quotative eṃdu is related to eṃn(v)- ‘say’ through its verbal paradigm. In its quotative use, it follows the material presented as spoken, thought, or otherwise framed as content. The citation form of the verb and the quotative can be compared with an inflected reporting predicate. The following table gives the forms relevant to that comparison (Kittel 1903, § on the verbal participles; Kittel 1894, § ಎನ್ನು).

FunctionFormTransliterationGlossTranslation
Lexical citation formಎನ್ನುennusay‘say’
Participial form used as a quotativeಎಂದುeṃdusay.CVB or QUOT‘having said’ or a marker following reported content
Finite past formಎಂದಂeṃdaṃsay.PST.3SG.M‘he said’

CVB identifies a converb, a dependent verbal form that can link a predication to another predicate. The lexical use ‘having said’ presents saying as part of a sequence of actions. The quotative use marks off content. The form alone can therefore leave two analyses available. Consider whether the clause introduces another action after a saying event, or whether a governing predicate takes the preceding material as its content. With a predicate of thought, the report can express an internal judgment without asserting any spoken utterance (Kittel 1903, § on the construction of participles; Kittel 1894, § ಎನ್ನು and § ನೆನೆ).

UseFormTransliterationGlossTranslation
Speech with a finite reporting verbಆನ್ ಬಂದೆಂ ಎಂದಂān baṃdeṃ eṃdaṃI.NOM come.PST.1SG say.PST.3SG.M‘He said, “I came.”’ (constructed)
Thought with a first-person perspectiveಆನ್ ಬಂದೆಂ ಎಂದು ನೆನೆದಂān baṃdeṃ eṃdu nenedaṃI.NOM come.PST.1SG QUOT think.PST.3SG.M‘He thought, “I came.”’ (constructed)
Quoted directiveಬಾ ಎಂದಂbā eṃdaṃcome.IMP.2SG say.PST.3SG.M‘He said, “Come.”’ (constructed)
Quoted nameವೀರ ಎಂದು ಕರೆದಂvīra eṃdu karedaṃVīra QUOT call.PST.3SG.M‘He called him Vīra.’ (constructed)

A reporting verb can itself introduce the quotation relation as in the first and third rows. The first-person ending in the first row is agreement inside the utterance, whose speaker is referred to in the third person by the reporting predicate. The understood second person of the imperative in the third row is the person addressed in the utterance. The quoted material can have the form and force of an independent utterance even though the passage presents it as the content of a speech event. A quoted name needs neither a subject nor a predicate. In the last row, the form before eṃdu supplies the name used in calling, and the person called is understood from context.

First-person forms, imperatives, and expressions of time or place whose reference depends on the speaker can help identify a direct perspective. In a third-person report, the same quotative may occur with a clause that English naturally renders with that. Person marking alone may leave the perspective undecidable, especially when neither the reported subject nor the reporting subject is overt. A passage can present a speaker referring to a third party, or report the content from the narrator’s perspective, with the same third-person morphology. The quotative identifies the content boundary more directly than it identifies the perspective.

The content normally precedes a following quotative. Its left boundary requires attention to meaning and syntax. A nominative immediately inside the content can belong to the reported predicate, while a nearby dative may belong to the reporting predicate as its addressee. A quotative after a word can enclose a name or term, and one after a finite predicate can close an entire clause. When the governing predicate follows, its own dependents are outside the content unless there is independent evidence that the quoted material includes them. Punctuation supplied in an edition or translation can indicate an editor’s reading. The inflections, the reporting predicate, and the shifts of perspective provide evidence in the wording itself.

Verbal sequences and auxiliary diagnostics

In a sequence with a non-finite verb followed by an inflected verb, the following verb can express a second lexical action, select the first action as its complement, or contribute a grammatical meaning to a complex predicate. Shared tense or a single overt subject can occur across these structures. The relation between the actions and the distribution of arguments distinguish them more clearly than the number of written words (Kittel 1903, § on the infinitive, verbal participles and auxiliary verbs).

SequenceFormTransliterationGlossTranslation
Two successive actionsಬಂದು ಕಂಡಂbaṃdu kaṃḍaṃcome.CVB see.PST.3SG.M‘He came and saw.’ (constructed)
An infinitival action and a predicate of beginningಮಾಡಲ್ ತೊಡಗಿದಂmāḍal toḍagidaṃdo.INF begin.PST.3SG.M‘He began to do it.’ (constructed)
A converb followed by ‘give’ಮಾಡಿ ಕೊಟ್ಟಂmāḍi koṭṭaṃmake.CVB give.PST.3SG.M‘He made it and gave it’ or ‘He made it for someone.’ (constructed)
A converb followed by ‘take’ಮಾಡಿ ಕೊಂಡಂmāḍi koṇḍaṃmake.CVB take.PST.3SG.M‘He made it and took it’ or ‘He made it for himself.’ (constructed)

In the first sequence, coming and seeing can be located as successive events. Seeing can have an object of its own. In the second, beginning applies to the doing, and the activity that begins is specified by the infinitive. The interpretations in the last two rows illustrate why ‘give’ and ‘take’ require attention to context. For the lexical readings, a thing made also undergoes transfer or taking. For the benefactive and self-directed readings, the contribution of the second verb concerns the participant for whom the making is undertaken. The table gives possible interpretations to distinguish in a passage. It does not establish an identical distribution for every first verb or every use of ‘give’ and ‘take’ (Kittel 1894, § ಕೊಡು and § ಕೊಳ್ಳು; Kittel 1903, § on auxiliary verbs).

Independent event and arguments. If the second verb retains a separately identifiable action, test whether its own object, recipient, goal, or other dependent can be recovered. In a literal transfer reading, there is something given and a recipient, even where the passage leaves one of them implicit. A benefactive interpretation can be supported by a context that provides a beneficiary and has only the making as the event under discussion. Shared participants alone permit both analyses. An object made and subsequently transferred can be understood with both verbs.

Scope and modification. A time expression that locates one action independently of the other can support a sequence. A tense or negative form interpreted over the linked predication can support a more closely integrated construction. Final inflection alone leaves scope open, since the last verb of a participial chain can carry the overt finite morphology. Use contexts that distinguish denial of the first action, denial of the second action, and denial of a single event with the relevant grammatical meaning. The interpretation depends on the placement of the negative expression and on what the surrounding passage asserts.

Recurrence and lexical contribution. An auxiliary analysis gains support when the same following verb contributes a stable grammatical meaning with first verbs describing different kinds of events. It gains further support if the following verb loses a requirement characteristic of its ordinary lexical use. A repeated sequence whose meaning depends on particular lexical items can instead be a fixed expression. A reading established with ‘make’ does not settle the interpretation after a motion, perception, or speech verb. Conversely, an item with an auxiliary use in one sequence can retain its lexical meaning elsewhere (Kittel 1903, § on auxiliary verbs).

A serial-verb analysis treats more than one lexical verb as part of a single predication. Evidence would include integrated event reference, shared arguments and scope, and a reason to analyze the verbs as a unit even though both retain a lexical contribution. A sequence with overt participial linkage and evidence for successive events has a straightforward clause-chain reading. A sequence with a following verb that expresses a regular grammatical contribution has a plausible auxiliary analysis. Where the text leaves the lexical and integrated readings both available, report the form of each verb and the scope that can be established. Adjacency, shared participants, and a natural one-verb translation are individually insufficient to establish serialization.

Evidence for control and raising analyses

Control describes a relation between an understood dependent subject and a participant licensed by the governing predicate. In a subject-control reading, the understood subject is identified with the governing subject. In an addressee-control reading, the addressee has a role in the governing speech event and is understood as the participant in the directed action. Raising describes a different relation. The expression associated with the dependent subject appears in a governing-clause position, but its participant role comes from the dependent predicate. These terms describe analyses of particular constructions. The inflection of an infinitive supplies no independent subject agreement with which to decide between them (Kittel 1903, § on the infinitive).

The directive ‘The king told the messenger to come’ gives the king and messenger roles in the telling. On the directive reading, the messenger is also the expected mover. This is evidence for an addressee-control analysis if the interpretation depends on the addressee role. Compare a finite message spoken to the same addressee about a third party. The addressee then continues to belong to the speech event, while the message supplies its own subject. Thus an addressee can be present whether or not it controls a dependent subject.

Form of the speech contentFormTransliterationGlossTranslation
Directed action with an understood actorದೂತಂಗೆ ಬರಲ್ ಪೇಳ್ದಂdūtaṃge baral pēḷdaṃmessenger.DAT come.INF tell.PST.3SG.M‘He told the messenger to come.’ (constructed)
Finite message with an expressed subjectಅರಸಂ ಬಂದಂ ಎಂದು ದೂತಂಗೆ ಪೇಳ್ದಂarasaṃ baṃdaṃ eṃdu dūtaṃge pēḷdaṃking.NOM come.PST.3SG.M QUOT messenger.DAT tell.PST.3SG.M‘He told the messenger that the king had come.’ (constructed)

The infinitival example with ‘begin’ is harder to classify from its surface pattern alone. In ‘He began to write’, a person both supplies the understood subject of ‘write’ and appears as the subject of ‘begin’. The construction permits an interpretation as the onset of the writing, but the example by itself leaves open whether this use of the governing predicate licenses a participant of its own. An animate subject performing a deliberate activity fits several analyses. Comparable uses with inanimate subjects or involuntary processes would test whether the restrictions on the subject come entirely from the dependent predicate.

EvidenceWhat to examineLimit of the test
A governing participant roleWhether the governing predicate independently requires a speaker, addressee, experiencer, or another participant, and whether that participant supplies the dependent subject.Coreference can also follow from a purpose relation or discourse continuity. Check that the dependent clause is selected.
Restrictions on the subjectWhether changing the dependent predicate changes the set of plausible subjects while the governing form stays the same. Compatibility with an inanimate subject or involuntary process can support an analysis in which the governing predicate applies to an event or proposition.Some control predicates allow involuntary action or inanimate participants. Establish the meaning of the governing predicate in the passage.
Dependents and caseWhether the candidate participant bears a case or has a position attributable to the governing construction while receiving its semantic role from the dependent predicate.Case can be licensed by the dependent predicate, and word order can vary. A nearby dative alone supplies no evidence for raising.
Independently identifiable subjectWhether a finite report or nominalized counterpart can express a different subject, and whether the infinitive in the same lexical construction permits that interpretation.A different form of complement can license a different construction. The finite counterpart does not settle the structure of an infinitive.
Idiomatic interpretationWhether a subject retains a meaning available only as part of an independently established idiom with the dependent verb.The idiomatic use must be established for Old Kannada and the same lexical sense. A modern translation cannot establish it.

For a raising analysis, there must be evidence both that the expression has governing-clause syntax and that the governing predicate contributes no independent role for it in that use. For a control analysis, there must be evidence for the participant role and for its identification with the understood dependent subject. A finite quotative clause with its own agreement and a distinct subject supplies direct evidence of an independently structured report. An infinitival or participial sequence with one overt participant leaves more of the relation to lexical meaning and context. The appropriate analysis can remain open for an individual occurrence when the passage provides no independent test.

How Mova handles complements and quotation

tools/mova-parse uses tools/langseg to assign languages to sentences and spans before parsing each span with its language model. Historical Kannada script falls under other. The Latin-script recognizers cover English and French, and the system also recognizes scholarly Hebrew or Aramaic transliteration. Where a span is assigned kn-x-old, there is no level-2 model for it. 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.

For kn-x-old, Mova stops before tokenization, tagging, lemmatization, and dependency parsing. It therefore supplies no Old Kannada tree that locates quotative boundaries, distinguishes infinitival complements from purpose clauses, links understood subjects, or analyzes verbal sequences. prag and world read English trees and have no parsed Old Kannada input. Mova holds no annotated Old Kannada treebank, so no treebank baseline is measured.

Squid reads the manual and turns its grammatical points into claims. Formalizable claims wait in its queue for a place in Mova’s code. The rules in this chapter remain unimplemented. The points that come first are those needed for a tokenizer, a tagger, and a Universal Dependencies treebank baseline.

Sources for this chapter

This page in Markdown https://md.mova.elamur.ai/free/en/docs/languages/kn/kn-x-old/22-complementation-and-quotation/