Mova0.6 pre-release

Lojban: an emblem of the logical language, for all and there is 13. Pro-arguments and pro-predicates

Lojban uses pro-arguments to identify speakers, point to things, recover referents, introduce variables and ask for missing information. A pro-argument in the class KOhA can form a sumti on its own. A form in GOhA can serve as the predicate or as a unit within a larger selbri. The grammar determines where each form can occur. Its interpretation can depend on the participants, the physical setting, an antecedent, an assignment or the scope of a quantifier. (Cowan 1997, §§2.3–2.4, 7; The Logical Language Group 1997, rules 95–97 and 130–152)

Speaker and addressee forms in the KOhA class

The speech situation supplies the referents of the following forms. In writing, the writer and intended reader can fill the speaker and addressee roles. The basic forms leave number unspecified. The group forms say which participant roles the group includes. (Cowan 1997, §§2.3, 7; The Logical Language Group 1994, cmavo list, KOhA)

FormGlossReferent and use
mispeakerThe speaker or those speaking together
doaddresseeThe person or people being addressed
mi'ospeaker and addresseeThe speaker and addressee together
mi'aspeaker and othersThe speaker together with others outside the addressee role
ma'aspeaker, addressee and othersThe speaker and addressee together with additional participants
do'oaddressee and othersThe addressee together with others outside the speaker role
koaddressee.IMPThe addressee, called on to make the predication true

A participant form can fill any argument place appropriate to its referent. The place structure of the predicate tells which participant does what. An occurrence of ko makes the predication a directive, and the place occupied by ko specifies the addressee's role in the requested situation. Several places can contain ko when the addressee is to fill each of them. A vocative can identify the person being addressed independently of the predicate's argument places. (Cowan 1997, §§2.3, 2.7, 2.14, 7; The Logical Language Group 1997, rules 40, 83, 97 and 415)

FormGlossTranslation
mi tavla dospeaker talk.to addresseeI talk to you. (constructed)
do tavla miaddressee talk.to speakerYou talk to me. (constructed)
ko tavla miaddressee.IMP talk.to speakerTalk to me. (constructed)
mi tavla kospeaker talk.to addressee.IMPLet me talk to you. (constructed)

In dialogue, a new speaker uses mi for their own turn and do for its addressee. In direct quotation, these forms are interpreted in the quoted speech situation. Repeating a predication with a pro-predicate can preserve its original participants across a change of speaker. See the last two sections for the form that updates those references. (Cowan 1997, §§7, 19)

Demonstratives and physical deixis

The three physical demonstratives select referents by relative distance in the speech situation. A gesture, direction of attention or shared view can identify the target. The distance distinctions are contextual. The middle form commonly points to something by the addressee when speaker and addressee occupy different positions. Each form can stand as a sumti, and the referent can be one thing or several. (Cowan 1997, §§2.3, 7; The Logical Language Group 1994, cmavo list, KOhA)

FormGlossRange
tithis.nearNear the speaker
tathat.medialAt an intermediate distance, often by the addressee
tuthat.farFarther away in the shared setting
FormGlossTranslation
ti cimoi le'i tcaduthis.near third.among described.set cityThis is third among the indicated set of cities. (excerpt, wiki:kacna'u)
ta plisethat.medial appleThat is an apple. (constructed)
tu cmanathat.far mountainThat over there is a mountain. (constructed)

The predicate supplies the classification in ta plise. To quantify over a demonstratively identified collection, place the quantifier before the demonstrative. Thus re ti has the gloss two this.near and means “two of these” when the gesture identifies a collection from which two can be taken (constructed). A description introduced by LE makes a sumti out of descriptive material, while the demonstrative itself supplies a sumti by pointing. The same outer quantifier position is available for both. (Cowan 1997, §§2.3, 2.10, 6–7; The Logical Language Group 1997, rules 95–97 and 300)

Assigned pro-arguments and anaphoric back-reference

Explicit assignments

The following ten forms are available as assignable argument labels. Their values come from assignments in the discourse. English translations can use a name, a noun phrase or a pronoun, according to the referent established for the label. (Cowan 1997, §7; The Logical Language Group 1994, cmavo list, KOhA)

SeriesFormsGloss
ko seriesko'a, ko'e, ko'i, ko'o, ko'uassigned argument 1–5
fo seriesfo'a, fo'e, fo'i, fo'o, fo'uassigned argument 6–10

With goi, an argument and an assignable form are placed in apposition to establish a shared referent. The description or name can come first, followed by goi and the label. The label can also come first, with the identifying term after goi. The construction is a relative phrase. Its terminator is ge'u, which can be omitted when the boundary is recoverable under the grammar. After assignment, an occurrence of the label can fill a different argument place and still pick out the same referent. (Cowan 1997, §§7–8; The Logical Language Group 1997, rules 95 and 121–122)

FormGlossTranslation
loi cmaci goi ko'adescribed.mass mathematics assign ARG.1mathematics, assigned to ko'a (excerpt, wiki:cmaci)
la solri goi ko'aname sun assign ARG.1the Sun, assigned to ko'a (excerpt, wiki:solri)
ko'a se mluni la lunraARG.1 swap.1.2 satellite.of name MoonIt has the Moon as a satellite. (excerpt, wiki:terdi)

In wiki:terdi, a preceding goi phrase assigns Earth to ko'a. The later occurrence of ko'a retains Earth as its referent, and se mluni puts the body orbited in the first place. The same label in wiki:solri has a separate assignment. An assignment can identify a group, an abstraction or other referent expressible by a sumti, as well as a named individual. Letter strings can also serve as argument labels. Syntactically, a letter string and a KOhA form are separate alternatives in the sumti rule. (Cowan 1997, §§7, 17; The Logical Language Group 1997, rules 97 and 817)

Back-reference without an assignment

FormGlossAntecedent
rilast.argumentThe most recent eligible completed sumti
rarecent.argumentA contextually identifiable recent sumti, usually farther back
ruremote.argumentA contextually identifiable more distant sumti

For ri, completion matters. When a description is still open, ri inside it can look back to an earlier completed argument. A completed sumti inside a larger construction can count. Earlier bare pro-arguments are generally passed over, with the physical demonstratives ti, ta and tu eligible as antecedents. The demonstrative's earlier use can include a gesture that established its referent. With ra and ru, distance helps narrow the search, and context identifies the intended antecedent. Their values depend on the discourse and have no fixed second- or third-argument index. (Cowan 1997, §7; The Logical Language Group 1994, cmavo list, KOhA)

FormGlossTranslation
mi viska lo gerku ku .i ri blabispeaker see described dog end.description SEP last.argument whiteI see a dog. It is white. (constructed)
la .alis. tavla lo pendo be riname Alice talk.to described friend link last.argumentAlice talks to a friend of hers. (constructed)

In the second example, ri fills the linked place in the predicate pendo. At that point lo pendo be ri is still being formed, and Alice is the completed antecedent. The boundaries of arguments therefore matter even when the relevant words are close together. (Cowan 1997, §7; The Logical Language Group 1997, rules 97, 111–112 and 150–161)

The following forms recover arguments by place. In the baseline account, they refer to the corresponding place of the main bridi of the sentence. Read the places from the predicate as it is used, including any conversion or explicit place tags. An embedded clause may have its own first argument while a vo'a within it refers to the first place of the main predication. (Cowan 1997, §§7, 9; The Logical Language Group 1994, cmavo list, KOhA)

FormGlossPlace recovered
vo'amain.x1First place
vo'emain.x2Second place
vo'imain.x3Third place
vo'omain.x4Fourth place
vo'umain.x5Fifth place

la .alis. viska vo'a has the gloss name Alice see main.x1 and means “Alice sees herself” (constructed). Here the reflexive reading follows directly from the first-place reference. The same mechanism can refer back to other places with the corresponding forms. (Cowan 1997, §7)

A relative clause's head

ke'a refers to the sumti that the relative clause is attached to. Within that clause, its position or place tag identifies the role of the head in the subordinate predication. Both restrictive poi and supplementary noi clauses can use it. When the relative role is recoverable, the clause can leave it implicit. An explicit ke'a is useful where several places are open or the intended place is later in the predicate's place structure. (Cowan 1997, §§7–8; The Logical Language Group 1997, rules 95, 97 and 121–122)

FormGlossTranslation
lo jmive zasti noi ke'a cu vasrudescribed living existence REL.SUPPL head PRED containliving existence, which contains … (excerpt, wiki:mivyske)
lo bangu poi ke'a se fintidescribed language REL.RESTR head swap.1.2 inventa language that is invented (excerpt, wiki:runbau)
lo gerku poi mi viska ke'adescribed dog REL.RESTR speaker see headthe dog that I see (constructed)

In the first two excerpts the head fills the first place of the predicate as written. In the last example it fills the second place of viska, and mi fills the first. The attachment of the relative clause supplies the referent in each case. (Cowan 1997, §8)

Missing arguments and explicit zo'e

A bridi can leave argument places unspoken. An unfilled place ordinarily has the interpretation available with explicit zo'e. The hearer can recover a value from context or leave the value unspecified to the degree the assertion permits. The predicate still has the place. A missing argument therefore supplies a possible site of contextual reference even when there is no overt pro-form. (Cowan 1997, §§2.3, 2.6–2.7, 7)

ExpressionGlossTreatment of the place
Unexpressed placeimplicit contextual valueThe place is understood from context or left unspecified
zo'eunspecifiedThe place is overtly occupied by a contextually supplied or unspecified value
zu'iusual.valueThe place is filled by the usual value for this relation in the circumstances
zi'odelete.placeThe asserted relation is understood with that place removed

In ordinary untagged order, an overt zo'e can occupy a place that would otherwise consume the next overt argument. This lets a later argument reach its intended place. A numbered place tag can instead address that later place directly. For a predicate cited on its own as a clause, the ordinary unexpressed places can all be understood contextually. In a description or relative clause, the enclosing construction can supply a role, as when the head is understood in an open place of the relative clause. (Cowan 1997, §§2.6–2.7, 7–9; The Logical Language Group 1997, rules 40, 53, 71 and 80–83)

FormGlossTranslation
mi tavla dospeaker talk.to addresseeI talk to you, with subject matter and language understood or unspecified. (constructed)
mi tavla do zo'e la .lojban.speaker talk.to addressee unspecified name LojbanI talk to you in Lojban about something understood or unspecified. (constructed)
mi tavla do fo la .lojban.speaker talk.to addressee place.4 name LojbanI talk to you in Lojban. (constructed)
mi tavla do zo'e zu'ispeaker talk.to addressee unspecified usual.valueI talk to you about something in the language usual in the circumstances. (constructed)

zi'o changes the relation asserted. For instance, mi zbasu lo dinju zi'o, glossed speaker make described building delete.place, asserts “I make a building,” using a relation with the materials place suppressed (constructed). The assertion leaves the question of materials outside that relation. With zo'e in the same position, the ordinary materials place remains and receives an unspecified value. (Cowan 1997, §7; The Logical Language Group 1994, cmavo list, KOhA)

An overt zo'e can also be a term in a larger sumti. The fragment .ezo'e in wiki:sanbau segments as .e zo'e, glossed and unspecified, and adds a contextually unspecified conjunct to the preceding list. Here the form supplies an argument for the connective as well as leaving its value open. (The Logical Language Group 1997, rules 92, 97 and 802)

Unspecified, indefinite and interrogative forms

Quantified and open arguments

zo'e leaves the value of a place to context or leaves its identification open. To introduce a bound argument variable, use da, de or di. The three spellings allow three variables to be distinguished within the same scope. They make the same range of entities available. With no expressed quantifier at introduction, the default is existential. Reusing a spelling under the same binding refers to the same variable. A prenex ending in zo'u can state the binding and its scope explicitly. (Cowan 1997, §§7, 16; The Logical Language Group 1997, rules 11, 30, 41 and 95–97)

FormGlossFunction
daVAR.1First argument variable
deVAR.2Second argument variable
diVAR.3Third argument variable
ce'uOPENOpen argument position in an abstraction
maARG.QArgument requested from the interlocutor

An overt quantifier can change the force of a variable. su'o da is existential, ro da universal and no da a zero quantification. A restrictive relative clause can delimit the range of the variable. Repetition of a bound form within that quantifier's scope preserves the binding. To carry a specific referent into successive independent statements, an assigned label makes that link explicit. (Cowan 1997, §§6–8, 16; The Logical Language Group 1997, rules 11–14, 30, 95–97 and 121–122)

FormGlossTranslation
so'idamany VAR.1many entities (excerpt, wiki:runbau)
ci klejutsithree taxonomic.classthree taxonomic classes (excerpt, wiki:finpe)
su'o da zo'u da tavla daat.least.one VAR.1 PRENEX VAR.1 talk.to VAR.1Someone talks to themself. (constructed)
ro da poi gerku cu danluevery VAR.1 REL.RESTR dog PRED animalEvery dog is an animal. (constructed)

Numerals and other quantifiers occupy a position outside a pro-argument or description, as in so'i da. A quantifier can also introduce a predicate directly, as in ci klejutsi. LE introduces a description, LA can introduce a name, and LI introduces a mathematical expression used as an argument. Thus a count in ci klejutsi constrains the number of classes, while the exhibit li 0 in wiki:kacna'u, glossed number zero, stands for the number zero. The baseline quantifier rule admits a number, with an elidable boi, or a mathematical expression introduced by vei, with an elidable ve'o. (Cowan 1997, §§6, 16, 18; The Logical Language Group 1997, rules 95–97, 300 and 812)

ce'u marks a place left open in the content of an abstraction, especially a property with ka. The place occupied by ce'u identifies the argument on which the property depends. Its contribution can lie within a clause embedded inside the abstraction. The extract lo nu pilno ce'u from wiki:lo jbobau has the gloss described event use OPEN and means “the event of using the entity in the open place.” In the containing ka abstraction, the open entity is the thing used. (Cowan 1997, §§7, 11)

Argument questions

ma asks what argument belongs in its position. The position, including any explicit tag, fixes the place being questioned. A reply can consist of the requested sumti, or can give a fuller predication containing it. With several occurrences, the question can request several values. The same form covers people, things and other kinds of arguments. For a polar question use the question indicator xu. To ask for the predicate use mo. (Cowan 1997, §§2.15, 7; The Logical Language Group 1997, rules 20, 83, 97 and 152)

FormGlossTranslation
ma tavla doARG.Q talk.to addresseeWho is talking to you? (constructed)
do tavla maaddressee talk.to ARG.QWhom are you talking to? (constructed)
do tavla fi maaddressee talk.to place.3 ARG.QWhat are you talking about? (constructed)
ma tavla maARG.Q talk.to ARG.QWho is talking to whom? (constructed)

References to utterances

An utterance can itself be the referent of a sumti. These forms locate it in the discourse. A form such as di'u picks out the earlier expression. With la'e di'u, la'e shifts reference to what that expression is about or denotes. This distinction matters when a following predicate is about wording, as opposed to the situation the wording describes. (Cowan 1997, §§2.12, 6–7; The Logical Language Group 1997, rule 97)

FormGlossUtterance referred to
da'uremote.past.utteranceA distant earlier utterance
de'urecent.past.utteranceA recent earlier utterance
di'uprevious.utteranceThe immediately preceding utterance
deicurrent.utteranceThe utterance in which the form occurs
di'enext.utteranceThe immediately following utterance
de'enear.future.utteranceA nearby forthcoming utterance
da'eremote.future.utteranceA more distant forthcoming utterance
do'iunspecified.utteranceAn utterance whose position is unspecified

The expression dei ci'arvelski lo tutra, glossed current.utterance written.description.of described territory, means “this utterance is a written description of the territory …” in wiki:auskalerik. The form dei identifies the text currently being read. A forward form such as di'e lets the speaker indicate the role of an utterance before giving it. (Cowan 1997, §7; The Logical Language Group 1994, cmavo list, KOhA)

Pro-predicates in the GOhA class

GOhA is a syntactic class. Its forms cover predication reference, variables over predicates, a predicate question and identity. Each can occupy the predicate position in a clause. It can also occur as a predicate unit in a larger selbri. In a description, the same syntax permits a predicate recovered from context to provide the descriptive content. (Cowan 1997, §§2.4, 7, 16; The Logical Language Group 1997, rules 53 and 130–152)

FormGlossUse
go'iprevious.bridiRepeat the immediately preceding predication
go'epenultimate.bridiRepeat the predication before the immediately preceding one
go'arecent.bridiRecover a recent more distant predication
go'uremote.bridiRecover a relatively distant earlier predication
go'ofuture.bridiRefer forward to a later predication
neicurrent.bridiRefer to the current predication
no'aenclosing.bridiRefer to the next enclosing predication
bu'aPRED.VAR.1First predicate variable
bu'ePRED.VAR.2Second predicate variable
bu'iPRED.VAR.3Third predicate variable
moPRED.QAsk for a predicate
duidentical.toAssert identity of the supplied referents

Reusing a predication

A backward reference form recovers the predicate and its place structure. For go'i used to repeat a bridi, its arguments are available to the new predication. Omitted places retain the recovered values. An overt argument replaces the value in the place it occupies, including when that argument is zo'e. Numbered place tags allow a later place to be replaced while earlier ones are retained. A bare go'i can confirm a polar question by repeating the questioned predication. (Cowan 1997, §§2.15, 7; The Logical Language Group 1997, rules 53, 71, 83 and 152)

The article wiki:xindo ropno bangu first uses cu bangu, glossed PRED language.of, in a statement with a lower bound on the number of speakers. It later ends a list of languages with lo dotco go'i, glossed described German previous.bridi, “[and] German likewise.” The list supplies the new first-place material, and the earlier predication gives the relation to the speaker population.

With the antecedent la .alis. tavla la .bob., glossed name Alice talk.to name Bob, “Alice talks to Bob,” the following continuations illustrate place replacement. (Cowan 1997, §7)

FormGlossTranslation
.i go'iSEP previous.bridiAlice talks to Bob. (constructed)
.i la .karol. go'iSEP name Carol previous.bridiCarol talks to Bob. (constructed)
.i go'i fe la .karol.SEP previous.bridi place.2 name CarolAlice talks to Carol. (constructed)
.i go'i fe zo'eSEP previous.bridi place.2 unspecifiedAlice talks to someone understood or left unspecified. (constructed)

The other reference forms use different ways of locating a predication. go'e counts back a fixed step in the sequence. go'a and go'u require an identifiable antecedent at a contextually appropriate distance. go'o requires a later predication to supply its content. nei and no'a locate predications by clause structure. no'a starts from an embedded bridi and goes to its containing bridi. A subscript can indicate a higher enclosing level. (Cowan 1997, §7; The Logical Language Group 1997, rules 32 and 152)

Variables, predicate questions and identity

The bu'a series allows predication itself to vary under a binding. The three forms can represent distinct predicate variables, and using one again within its binding recovers the same predicate value. With mo, the interlocutor supplies a predicate that fits the clause, with the expressed arguments already specifying places. With du, the speaker says the terms designate the same entity or entities. A predication that merely classifies an entity can use the classifying predicate directly. (Cowan 1997, §§2.4, 2.15, 7, 16)

FormGlossTranslation
do moaddressee PRED.QWhat are you doing? / What are you? (constructed, interpretation depends on context)
ke'a du la merkuris.head identical.to name Mercurywhich is Mercury itself (excerpt, wiki:solri)

The baseline licenses ra'o immediately after GOhA in a predicate unit. On a pro-predicate that recovers a predication, ra'o directs the hearer to update the interpretation of its inherited pro-arguments to the current context. Its effect is visible in dialogue where the speaker and addressee exchange roles. (Cowan 1997, §7; The Logical Language Group 1997, rule 152; Hagiwara, Ilmen and lagleki 2014, rules GOhA and RAhO)

Resetting assignments and maintaining reference across utterances

da'o cancels the current assignments of pro-arguments and pro-predicates. After a cancellation, an assignable form can be established again with goi, and a fresh pro-predicate assignment can use cei. The baseline admits da'o in the indicator construction. It admits cei between predicate units in the construction that establishes a predicate assignment. Participant forms still take their referents from the speech situation, and the scope of a quantified variable comes from its binding. (Cowan 1997, §§7, 16; The Logical Language Group 1997, rules 150–152 and 411–413; Hagiwara, Ilmen and lagleki 2014, rule DAhO)

FormGlossTranslation
lo gerku cu blabi cei bu'adescribed dog PRED white assign.PRED PRED.VAR.1The dog is white, with bu'a assigned to the predicate ‘be white’. (constructed)
da'oclear.assignmentsClear the current pro-form assignments. (constructed, indicator shown on its own)

An ordinary .i separates statements. An assignment can support later reference across that boundary, as in wiki:mivyske, where lo mivyske goi ko'a, glossed described biology assign ARG.1, assigns biology to ko'a. The later .i ko'a skegau, glossed SEP ARG.1 science.do, begins a further predication about biology. The intervening relative clause contains ke'a, whose referent comes from its relative attachment. The later ko'a recovers its referent from the earlier assignment. (Cowan 1997, §§2.8, 7–8; The Logical Language Group 1997, rules 2–14 and 121–122)

Topic markers give the hearer information about which stretch of discourse is current. ni'o introduces a new topic and no'i resumes an earlier topic. In wiki:sanbau, ni'o lo sanbau goi ko'a, glossed new.topic described Spanish.language assign ARG.1, establishes a topic and assigns it a label. The later ko'a in the same stretch of text keeps that referent. A topic marker organizes the text, while goi supplies a specific identity for the label and da'o explicitly clears assignments. (Cowan 1997, §§7, 19; The Logical Language Group 1994, cmavo list, NIhO and DAhO; The Logical Language Group 1997, rules 2–14 and 413)

Across turns, a bare repetition can keep the participants of the earlier predication even though the current mi and do have changed. An explicit replacement in a place uses the expression supplied in the new turn. ra'o instead updates inherited pro-arguments against the new context. In the constructed exchange below, A addresses B. The two B rows are alternative replies to A. (Cowan 1997, §7)

TurnFormGlossTranslation
Ami prami dospeaker love addresseeI love you.
Bgo'iprevious.bridiYes, you love me.
Bgo'i ra'oprevious.bridi update.referencesI love you.

The baseline EBNF provides a position for KOhA in a sumti, GOhA with optional RAhO in a predicate unit, and DAhO among indicators. The PEG grammar also recognizes these classes and provides a KOhA clause with pre- and post-clause components. These provisions identify the constructions to which reference interpretation applies. Resolving a demonstrative, an assignment or a discourse antecedent also requires the corresponding information from the utterance and its context. (The Logical Language Group 1997, rules 97, 152 and 413; Hagiwara, Ilmen and lagleki 2014, rules KOhA_clause, GOhA, RAhO and DAhO)

How Mova handles Lojban pro-forms and coreference

Mova has no level-2 parser model for jbo. tools/langseg has neither a Lojban lexicon nor a Lojban language code, so it cannot distinguish Lojban spans from other Latin-script text. For a span without a model, tools/mova-parse reports the missing parse in the # parse comment. The tokens receive Parse=None in MISC and _ in LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL.

Mova has no Lojban word breaker, parser or treebank. The research library holds the 1997 YACC and EBNF grammar, the ilmentufa camxes.peg grammar, and the official gismu, cmavo, rafsi and lujvo lists. These materials are available in the library, and Mova's code has yet to use them. There is no measured Lojban treebank baseline or mapping from Lojban parse trees to Universal Dependencies, frames or pragmatics. prag and world read English trees and have no Lojban trees to process.

Squid reads this chapter as grammatical claims. Formalizable claims wait in its queue for implementation. Mova has yet to implement the chapter's rules for recognizing Lojban pro-forms, interpreting omitted arguments or resolving assignments and antecedents. The first dependencies are 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/jbo/13-pro-forms-and-reference/