Mova0.6 pre-release

Lojban: an emblem of the logical language, for all and there is 11. Place conversion, place tags and valency

A Lojban selbri supplies the relation in a clause, and its sumti supply arguments of that relation. Each numbered place has a role, such as runner, running surface or gait. Conversion changes which role has which number. Place tags assign arguments to those numbers, and modal tags relate further arguments to the clause. Causation and relations involving the same participants in several places can also be expressed through the choice of predicate or argument.

Reordering predicate places with the SE class

A predicate word can be a root (gismu), a compound (lujvo) or a borrowing (fu'ivla). Its lexical entry supplies the ordered places. The category of the word by itself gives neither the number of places nor their roles. For example, bacru has an utterer and an utterance, while bajra has the following four places. An audience for bacru would require separate expression, since its second place is what is uttered. (The Logical Language Group 1994, §§ bacru, bajra; The Logical Language Group 1997, § 152)

PlaceRole in bajra
x1Runner
x2Surface traversed
x3Limbs used
x4Gait

An SE word before a predicate unit exchanges the first place of that unit with one other place. The remaining places keep their numbers. The identity order is expressed with the bare predicate. A conversion needs the place it selects to exist. For instance, xe bajra has no fifth lexical place to exchange. The syntax of an SE prefix by itself leaves this check to the interpretation of the predicate. (The Logical Language Group 1994, §§ se, te, ve, xe, bajra; The Logical Language Group 1997, § 152)

FormPlace exchangeOriginal roles in the resulting order, for a five-place predicate
Bare predicateIdentityx1, x2, x3, x4, x5
seFirst and secondx2, x1, x3, x4, x5
teFirst and thirdx3, x2, x1, x4, x5
veFirst and fourthx4, x2, x3, x1, x5
xeFirst and fifthx5, x2, x3, x4, x1

The table is the complete SE paradigm for predicate places. It applies to shorter place structures wherever the selected place exists. SE can prefix another converted unit, so a series of prefixes is interpreted from the innermost outward. With se te bajra, te first gives the order limbs, surface, runner, gait. The outer se gives surface, limbs, runner, gait. Two occurrences of the same SE word in immediate succession restore the input order. (The Logical Language Group 1997, § 152)

FormGlossTranslation
mi bajra le darguI run DESC roadI run along the road.
le dargu cu se bajra miDESC road PRED CONV.2 run IThe road is run along by me.
lo tuple cu te bajra le dargu miDESC leg PRED CONV.3 run DESC road IThe legs are what I use to run along the road.

All three examples are constructed. DESC marks a description, PRED marks the boundary before a selbri, and CONV.2 and CONV.3 mark exchanges with the second and third places. The road stays the surface in all three clauses. In the last example the runner is third, while the surface is still second.

In a tanru, a sequence of predicate units, the final head normally supplies the places of the whole, with earlier material restricting its interpretation. SE immediately before a word converts that unit. bo binds neighboring material more tightly than ordinary juxtaposition. ke and ke'e delimit a group that can itself be converted. With co, the head precedes the modifier. Establish this grouping before deciding what any SE prefix converts. When a unit has arguments attached with be, subsequent linked terms use bei; be'o closes the links and can be elided when the boundary is recoverable. A link to a converted unit uses that unit's converted order. (The Logical Language Group 1997, §§ 130–136, 150–152, 160–161)

FormGlossInterpretation
barda se bajralarge CONV.2 runA running-surface predicate qualified by largeness.
se barda bo bajraCONV.2 large TIGHT runA runner predicate modified by converted barda, whose first place is a dimension of largeness.
se ke barda bajra ke'eCONV.2 GROUP large run END.GROUPThe second place of the grouped barda bajra relation becomes first.

These are constructed predicate phrases. The glosses leave the semantic link between the modifier and the head open. The grouping determines which places are exchanged even where the modifier could be understood in several ways.

Assigning an argument to a numbered place with the FA class

In the ordinary clause pattern, untagged sumti before the selbri start at x1 and proceed in order. Untagged sumti after the selbri continue the sequence. If the clause starts with the selbri and has supplied no earlier positional term, the untagged series after it starts at x2, leaving x1 implicit. An explicit zo'e fills a place with a contextually understood or unspecified value and counts in the sequence. cu, when present, separates the preceding terms from the selbri. (Cowan 1997, §§ 9.2–9.3; The Logical Language Group 1997, §§ 40, 53, 71, 80–83)

A numbered FA tag assigns the following sumti directly to the indicated place of the current selbri. This allows a later place to appear early or an earlier place to appear after the selbri. An untagged sumti following a numbered assignment continues at the next higher place, skipping places already filled. Another numbered tag establishes a new starting point. The same numbered tags can be used on linked terms within a predicate unit. Their numbers are local to that unit in its current form. (Cowan 1997, § 9.3; The Logical Language Group 1997, §§ 83, 151, 160–161)

FA formAssignment
faFirst place
feSecond place
fiThird place
foFourth place
fuFifth place
faiSpecial place for the original first argument in a JAI construction
fi'aQuestion asking which place the tagged argument fills

fai and fi'a are members of FA with special functions. fai addresses the displaced argument, and fi'a leaves the place number open for the answer. An FA term can also use ku in lieu of an overt sumti. The resulting term has the specified place assignment with its value left implicit. The baseline grammar licenses elision of a terminator where the parse stays unambiguous. (The Logical Language Group 1994, §§ fa, fe, fi, fo, fu, fai, fi'a; The Logical Language Group 1997, § 83)

FormGlossTranslation
fe le dargu cu bajra fa miPLACE.2 DESC road PRED run PLACE.1 II run along the road.
bajra fi lo tuple lo cadzu tadji fa mirun PLACE.3 DESC leg DESC walk manner PLACE.1 II run on my legs with a walking gait.
se bajra fa le dargu fe miCONV.2 run PLACE.1 DESC road PLACE.2 II run along the road.
do cu bajra fi'a le darguyou PRED run WHICH.PLACE DESC roadWhat role does the road have in your running?

These examples are constructed. In the second, fi assigns the limbs to x3, so the untagged manner goes in x4. The surface remains implicit. In the third, the road is x1 of se bajra and the runner is x2. A speaker who answers the fourth with fe identifies the road as the running surface.

The corpus has “se vacri fi loi trano” (sentence ID wiki:terdi). Its gloss is CONV.2 atmosphere PLACE.3 MASS nitrogen, and the phrase in context means “has an atmosphere containing nitrogen.” Earth is the first argument of se vacri. The atmosphere is implicit in the second place, and fi supplies a component of its composition in the third. This conversion exchanges only the first two roles.

Distinguishing lexical places from added modal terms

A lexical place comes from the selbri's place structure. A term introduced by a modal tag relates its sumti to the clause by an additional relation. The added term consumes no numbered place, so the next untagged positional sumti continues the numbered sequence. This is useful even where a different predicate already has a lexical place for the same general kind of information, such as a language, a source or a standard. Time and space tags can also introduce tagged terms. The grammar places modal or tense tags and FA in separate alternatives for forming a term. (Cowan 1997, §§ 9.5–9.6; The Logical Language Group 1997, §§ 83, 491, 815, 972)

A BAI modal typically takes its relation from a lexical predicate. The plain tag uses that predicate's first place for the modal sumti. SE before BAI selects another place of the source predicate, where that place exists. In se bau, the selected place belongs to the predicate behind bau. No place of the main selbri changes number. (The Logical Language Group 1994, §§ bau, bangu, bai, bapli, be'i; The Logical Language Group 1997, §§ 491, 815, 972)

TagSource predicateRole of the attached sumti
baubanguLanguage used, bangu x1
se baubanguUser of the language, bangu x2
te baubanguContent expressed, bangu x3
baibapliCompelling force, bapli x1
be'ibenjiSender, benji x1

These are examples of BAI usage. A productive way to form a modal tag is fi'o plus a selbri and the elidable terminator fe'u. The modal sumti fills the first place of the selbri inside the tag. To select another place there, convert that inner selbri. fe'u closes the tag; the associated sumti follows it. Where the end of the inner selbri is clear, fe'u can be left out. (Cowan 1997, § 9.5; The Logical Language Group 1997, §§ 83, 130, 491, 815)

FormGlossTranslation
mi bacru bau la .lojban. zo coiI utter IN.LANGUAGE NAME Lojban WORD helloI say the word coi in Lojban.
mi bacru fi'o bangu fe'u la .lojban. zo coiI utter MOD language END.MOD NAME Lojban WORD helloI say the word coi in Lojban.

Both examples are constructed. The name after the tag supplies the language. The quoted word is the utterance in bacru x2, although it comes after the language term. The corpus has “fi'o kakne le nu jmina je galfi sinxa” (sentence ID wiki:bangu). Its gloss is MOD capable DESC EVENT add AND modify sign, roughly “with the ability to add and modify signs.” fi'o kakne forms the tag, and the activity description follows with fe'u elided.

The JAI construction and its retained argument

jai prefixes a predicate unit and can take a tag of its own. With a tag, it gives the role associated with that tag the first place of the resulting predicate. The old first role is available through fai, and the old second and later numbered roles keep their numbers. The tag between jai and the unit is part of the conversion. The sumti that supplies this role can occupy the usual first-place position or be assigned there explicitly with fa. A fai term can express the displaced first argument wherever the term is licensed. That argument can also be left implicit. (The Logical Language Group 1994, §§ jai, fai; The Logical Language Group 1997, §§ 83, 152, 491, 815)

RoleWith an ordinary bau term on bacruWith jai bau bacru
Uttererx1fai
Utterancex2x2
Languagebau termx1
FormGlossTranslation
mi bacru zo coi bau la .lojban.I utter WORD hello IN.LANGUAGE NAME LojbanI say the word coi in Lojban.
la .lojban. jai bau bacru zo coi fai miNAME Lojban JAI IN.LANGUAGE utter WORD hello DISPLACED ILojban is the language in which I say the word coi.

The examples are constructed. In the second, the language is the ordinary first argument. fai mi supplies the speaker and zo coi is still second. This is the construction closest to an applicative when the effect sought is to give a modal participant a numbered core place. The choice of tag determines its role, so the construction can be used for different relations.

Without a tag, jai has an abstraction-raising use. A sumti can stand in the first place for something associated with the original first-place abstraction. The association is supplied by context. An object or a person in that position can, for example, suggest something it does or an event involving it. The underlying abstraction can be supplied with fai. With an unconverted predicate, tu'a provides another way to refer indirectly to an abstraction involving a sumti. The bare jai construction leaves later places available in their usual order. (The Logical Language Group 1994, §§ jai, fai, tu'a; The Logical Language Group 1997, §§ 83, 97, 152)

FormGlossTranslation
lo nu mi bacru cu cfipu doDESC EVENT I utter PRED confuse youMy speaking confuses you.
mi jai cfipu doI JAI confuse youSomething involving me confuses you.
mi jai cfipu do fai lo nu mi bacruI JAI confuse you DISPLACED DESC EVENT I utterI confuse you by speaking.

These examples are constructed. The final clause specifies the abstraction that the shorter raised clause leaves to context. In all three, the person experiencing confusion is the second argument.

Causative predicates and derived place structures

Lojban can express causation with a separate predicate whose places include the cause or agent and the resulting event. An abstraction introduced by nu allows the resulting event to be supplied as a sumti, with its own predicate and argument assignments. The choice of causative predicate matters. An intentional agent, an event that causes a result, and a force that compels a result have different lexical descriptions. (The Logical Language Group 1994, §§ gasnu, rinka, bapli; The Logical Language Group 1997, § 152)

PredicateFirst placeSecond placeFurther lexical place
gasnuAgent who does or brings about somethingAction or event brought about—
rinkaCausal event or stateResulting event or stateConditions under which the cause operates, x3
bapliCompelling forceEvent compelled to occur, or property made manifest—

bapli describes successful compulsion. An attempt to compel can be expressed by putting the compelling event inside an attempt predicate. gasnu makes the actor responsible for bringing the event about. The participants of that event are assigned inside the abstraction, independently of the assignment of the outer agent. Thus in the first example below, the first place of gasnu is filled by mi, while the first place of bajra is filled by do. (The Logical Language Group 1994, §§ bapli, gasnu, bajra, troci; The Logical Language Group 1997, §§ 41, 152)

FormGlossTranslation
mi gasnu lo nu do bajra le darguI bring.about DESC EVENT you run DESC roadI make you run along the road.
lo nu mi bacru cu rinka lo nu do bajraDESC EVENT I utter PRED cause DESC EVENT you runMy speaking causes you to run.
mi troci lo nu gasnu lo nu do bajraI try DESC EVENT bring.about DESC EVENT you runI try to make you run.

These examples are constructed. The last asserts the attempt and places the causation within its scope.

A causative can also be built into a compound. gau, a combining form of gasnu, is used in compounds for making or doing something. ri'a, a combining form of rinka, is available for compounds based on causation. A causative compound can have a first place for the causer and further places for participants of the result. The place structure belongs to the resulting lexical predicate. The combining forms help identify the relation, while the entry for the compound determines which component places remain, which are identified with one another and where the retained roles occur. A speaker introducing a new compound can specify its intended places when the formation leaves them unclear. Once the order is established, SE and FA operate on that order. (The Logical Language Group 1994, §§ gasnu, gau, rinka, ri'a, lujvo list; The Logical Language Group 1997, §§ 83, 152)

The corpus contains “le nu porgau” (sentence ID wiki:kacna'u). The gloss is DESC EVENT put.in.order, and the fragment refers to putting things in order. Compare the following constructed illustration. Here bajgau is used with causer first, runner second and surface third.

FormGlossTranslation
mi bajgau do le darguI cause.to.run you DESC roadI make you run along the road.
do se bajgau mi le darguyou CONV.2 cause.to.run I DESC roadYou are made to run along the road by me.

The simple gasnu construction keeps a whole event description in one place of the causative. In the illustration with bajgau, the runner and surface are separate places of the compound, so conversion of that compound can select the runner directly.

Reflexivity and reciprocity in argument expressions

A reflexive reading arises when two places are filled by expressions that refer to the same participant. The vo'a series provides pro-sumti that refer by place number to an argument of the main bridi. In a simple clause, identify the numbered argument after conversion and explicit place assignment, then interpret the pro-sumti by that number. Reflexive reference supplies a value for a place and leaves both role positions in the predicate's structure. (Cowan 1997, § 7.6; The Logical Language Group 1994, §§ vo'a, vo'e, vo'i, vo'o, vo'u; The Logical Language Group 1997, § 97)

Pro-sumtiAntecedent
vo'aFirst argument
vo'eSecond argument
vo'iThird argument
vo'oFourth argument
vo'uFifth argument
FormGlossTranslation
mi viska vo'aI see REFL.1I see myself.
mi se viska vo'aI CONV.2 see REFL.1I am seen by myself.
mi viska miI see II see myself.

All three examples are constructed. In the first, the antecedent is first as the viewer. In the second, it is first as the person viewed, and the viewer is second. The last example uses ordinary repeated reference to fill both places.

A reciprocal expresses the relation with the participants' roles reversed as well. The free marker soi takes one or two sumti and can be closed by se'u. With two sumti, it identifies both arguments whose places are exchanged in the additional assertion. With one sumti, the other is supplied by the sumti immediately before the marker. The interchange concerns the indicated roles of the same relation. Other arguments, when supplied, remain in their roles. This makes the reciprocal more specific than the mere co-occurrence of two people in a clause. (Cowan 1997, § 7.6; The Logical Language Group 1994, §§ soi, se'u; Hagiwara, Ilmen and lagleki 2014, § SOI_clause)

FormGlossTranslation
mi viska do soi vo'aI see you RECIP REFL.1You and I see each other.
mi viska do soi mi do se'uI see you RECIP I you END.RECIPYou and I see each other.
mi viska do .i je do viska miI see you SENT AND you see II see you, and you see me.

These are constructed. The one-sumti soi form pairs the preceding do with vo'a, which refers to mi. The form with two overt sumti names the pair directly. The final example states the two directions separately, a pattern that also lets a speaker change time, location or another argument from one direction to the other.

Effects of conversion on descriptions and clause interpretation

A description built directly on a selbri draws its referent from that selbri's first place. Converting the selbri inside the description therefore changes the role from which the referent is taken. The description as a whole can then occupy any suitable place of an outer predicate. That outer assignment and the inner first place are independent. A be link supplies an argument inside the description, ordinarily starting at the current second place. FA can direct the linked sumti to a particular numbered place. (The Logical Language Group 1994, §§ be, bei, be'o; The Logical Language Group 1997, §§ 83, 97, 111–112, 151–152, 160–161)

DescriptionGlossReferent
lo bajraDESC runA runner
lo se bajraDESC CONV.2 runA surface someone runs on
lo te bajraDESC CONV.3 runLimbs used in running
lo ve bajraDESC CONV.4 runA gait used in running
lo se bajra be mi be'oDESC CONV.2 run LINK I END.LINKA surface I run on

These are constructed examples. In the last, be mi supplies x2 of se bajra, the runner. In the corpus, “lo se lifri” (sentence ID wiki:bangu) has the gloss DESC CONV.2 experience and means “what is experienced.” The experience occupies the first place of the converted selbri inside the description, wherever the resulting sumti is used.

The passive is a frequent translation of se when a predicate's original second argument is patient-like. In such examples, the original first argument stays available in the second place. It can be expressed overtly, given explicitly as zo'e after conversion or left implicit. The same exchange works with a surface or a dimension, and the other SE members can put a third, fourth or fifth argument first. The interpretation follows the lexical roles and their assignment. (Cowan 1997, § 9.4; The Logical Language Group 1994, §§ se, te, ve, xe, barda, bajra; The Logical Language Group 1997, §§ 83, 152)

FormGlossTranslation
mi viska le mlatuI see DESC catI see the cat.
le mlatu cu se viska miDESC cat PRED CONV.2 see IThe cat is seen by me.
le mlatu cu se viskaDESC cat PRED CONV.2 seeThe cat is seen.
mi viskaI seeI see something.

These are constructed. The first two express the same relation with both participants identified. The third leaves the viewer implicit. The fourth keeps the viewer first and leaves what is seen implicit. When translating an antipassive-like clause with a backgrounded patient, ordinary omission can give the fourth pattern. The omitted role is still part of the lexical place structure.

Middle readings depend on what the clause means in context. A disposition of a participant can be expressed by a predicate of ability or ease with an embedded event. A self-directed event can be expressed by reflexive reference. If an event is described with an implicit actor, that actor's role belongs to the chosen predicate. Neither the availability of a passive translation nor the omission of an argument establishes a distinct middle or antipassive place paradigm. SE exchanges existing places, FA assigns them, and JAI can give first place to an added tagged role or to a sumti associated with an abstraction. (The Logical Language Group 1994, §§ frili, kakne, vo'a, jai; The Logical Language Group 1997, §§ 83, 97, 152)

For a clause with a complex selbri, establish the predicate grouping and the places of its head, then apply any conversion within its scope. Assign the overt sumti by tags or by their position. Resolve a reflexive by the resulting place numbers and interpret a reciprocal by the indicated pair of roles. With an abstraction, each embedded predicate has its own place structure. These dependencies account for why a change in grouping, a conversion inside a description and a change in the order of explicitly tagged arguments can have different effects on the interpretation. (The Logical Language Group 1997, §§ 40–41, 83, 97, 130–136, 150–152, 160–161)

How Mova handles place conversion and frame roles

Mova has no Lojban word breaker, parser model or treebank. tools/langseg has neither a Lojban lexicon nor a jbo language code, so it cannot identify Lojban spans among other Latin-script text. For tokens of jbo without a model, tools/mova-parse leaves lemma, part-of-speech, feature and dependency fields unannotated and records Parse=None in MISC.

The research library holds the 1997 YACC and EBNF baseline grammar, the PEG grammar camxes.peg, and the official gismu, cmavo, rafsi and lujvo lists. These materials are available as references. Mova's code currently uses none of them. There is no measured Lojban treebank baseline and no Lojban mapping to Universal Dependencies, frames or pragmatics. prag and world read English trees.

Squid reads the manual as grammatical claims and queues the formalizable ones. The place exchanges, explicit place assignments, modal terms, displaced arguments and argument relations described here have no implementation in Mova yet. The initial work would be the word segmentation, tagging and treebank baseline needed to obtain Lojban trees.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/jbo/11-place-conversion-and-valency/