Skip to content

Latest commit

 

History

History
354 lines (249 loc) · 17.2 KB

verbal-constructions.md

File metadata and controls

354 lines (249 loc) · 17.2 KB
Raw

Specific verbal constructions and their rolesets

UMR, following AMR, represents some verbal constructions in a different way than we are used to in PDT-like annotations.

  • PDT distinguish lexical verbs (= autosemantic verbs as action verbs like jít, dělat, snídat, měnit but also phase verbs like začít, skončit combined with infinite verbs (Začal pracovat.) and copula být (Petr je doktor / chytrý., Jan bývá často nemocný.)) represented as t-nodes and auxiliary verbs (= synsemantic units, esp. auxiliary být, but also modal verbs like muset, chtít) represented as attributes of respective t-nodes.

  • In UMR, there are also two types of "predicates" in UMR -- those headed by "lexical" predicates (corresponding to particular verbs and their rolesets, as defined in PropBank) and those referred to as "abstract rolesets" of different types, which are meant to annotate the crosslinguistically stable meanings.

Also the distinction what should be considered a single event and what are two events (annotated as two verbs) differ

Autosemantic být (incl. copula verbs)

PDT distinguishes several types of constructions with být "to be":

  1. existential být,
  2. substitute být,
  3. copula být,
  4. phrasal být, and
  5. být in constructions with a single constituent.

All of them are treated as lexical verbs in PDT and they should also be treated as a (single) eventive concepts in UMR.

1. Existential být

Existential být is designated as meaning that "something is/is not, exists/does not exist", as:

  • [cs] Strašidla.ACT na světě nejsou.
    "There are no ghosts in the world." (PDT manual).

It should correspond to být-011 in PDT-Vallex.

See also existential be-02 in PropBank.

In UMR, thetic location exist-91 or presentational location have-place-91 are probably the best fit for the existential být.

  • [cs] V pohádkách jsou různá strašidla.ACT. ... ?? exist-91 (thetic location)
  • [cs] Strašidla.ACT jsou jen v pohádkách. ... ?? have-place-91 (predicative location)

2. Substitute být

Substitute být stands for some full verb which can be substituted for it;

  • Typically, it is characterized by a valency frame with just ACT, possibly complemented with an optional free modification, as:
    Jirka je na zahradě.LOC "George is in the garden." (= George is located/appears in the garden)

This type should correspond to být-011 in PDT-Vallex (i.e., existential and substitute být is not distinguished in PDT-Vallex).

As for UMR, different types of UMR abstract predicates, other -91 substitutes, or even full verb predicates should be used.

  • [cs] Jirka je na zahradě.LOC "George is in the garden." ... ?? have-place-91 (predicative location)
  • [cs] Úkol byl na pátek.TOWH "The assignment was for Friday." ...???
  • [cs] To je pro mě.BEN "That is for me." ... ???
  • [cs] Je pozdě. TWHEN cokoliv dělat. "It is too late to do anything." ...???
  • [cs] Vystoupení bude bez ohledu na počasí.REG "The performance will take place regardless of the weather." ...???
  • [cs] Zájezd byl prostřednictvím kanceláře.MEANS "The excursion was arranged by the office." ... ???
  • [cs] To bylo schválně.CAUS "That was on purpose." ...???
  • [cs] To bylo o chlup.DIFF "That was close." ...???
  • [cs] Chaloupka je jako dlaň.CPR "The cottage is tiny." ... have-mod-91
  • [cs] Tento nástroj je na stáčení. AIM vína. "This instrument is for bottling wine." ...???
  • [cs] Byli jsme to obhlížet.INTT "We have been to inspect it." ...???
  • [cs] Zahrada je souseda.APP "The garden is the neighbour's." ...???

Rarely, valency frames with two (or more) arguments come into play, as in:

  • [cs] Rukavice.ACT mu.PAT nejsou. "The gloves do not fit him." ...???

3. Copula verb

Copula verb as a part of verbonominal predicate, i.e. predicate constituted of:

  • a verbal part (být, bývat, stát se, stávat se, as Stal se učitelem."He became a teacher." and

  • a non-verbal part

    • typically a semantic adjective or a noun in the nominative or the instrumental, as:
      Kočka je savec.PAT. "The cat is a mammal." --> have-mod-91
      Jirka je hodný.PAT. "George is good." have-mod-91
      Dětí je pět.PAT "There are five children." --> have-quant-91
    • also a noun in the genitive, an infinitive, a dependent clause, an adverb or an interjection, as Je vidět.PAT Sněžka. "Sněžka can be seen." --> ??
      Končit není umřít.PAT "To finish is not to die." --> ?? _Jeho výklad je, že zahrají.PAT_ "According to him, they will play." _To je moc.PAT_ "That is too much." --> have-quant-91 _To je fuk.PAT_ "That doesn't matter. --> ??`
    • a subtype with a noun or an adjective with modal meaning are also treated verbonominal predicates, like:
      být schopný.PAT (=to be capable), být možné.PAT (=to be possible), být povinností.PAT (=to be an obligation) --> ??

Copula verb být corresponds to the být-007 roleset (with 2 arguments, ACT and PAT).

See also copula be-01 in PropBank.

According to the UD guidelines:

The copula verb být "be" is used in equational, attributional, locative, possessive and benefactory nonverbal clauses. Purely existential clauses (without indicating location) use být as well but it is treated as the head of the clause and tagged VERB.

The copula být should correspond to být-007 in PDT-Vallex. However, this frame is assigned also to constructions with location meaning (e.g., Jiřinka je tady, ta s tou tmavší mašlí `Jiřinka is here, the one with the darker bow' [= a comment on the photo]).

4. Phraseological být

These constructions are characterized by the DPHR functor in its valency frame.

5. Být in constructions with a single constituent

  • impersonal constructions ... být-017 je horko "It is hot." --> ??
  • constructions of experience (with ACTor in dative) ... být-009 je mi horko --> ??

Temporal and modal auxiliaries - a single event

Czech is an inflective language so analytical verb forms are not so frequent as in English. However, auxiliaries are still used to create some verb forms, as simple past tense (just 1st and 2nd person) Přišel jsem. "I came." or future tense Budu teprve obědvat. "I'll have lunch only now." (imperfective verbs).

Temporal auxiliaries

In PDT, the following verbs are treated as (temporal) auxiliaries:

  • být "be"
  • in diatheses, also mít "have" (resultative diathesis) and dostat "get" (recipient diathesis) are used.

As a consequence, these verbs are not treated as separate t-nodes - instead, they contribute to values of selected grammatemes (esp. the grammatemes of tense, diatgram, and factual modality (condition).

The treatment of temporal auxiliaries in UMR is similar: they are not identified independently as events - instead, they are treated within temporal dependency annotation, as in She has gone/is going to school. (treated as on going event).

Modal auxiliaries

Modal auxiliaries in PDT

In PDT, modal auxiliaries are covered by the deontmod grammateme with the following values:

  • deb (debitive) the event is understood as "necessary" ... muset "must, have to",
  • hrt (hortative) the event is understood as "obligatory (an obligation)" ... mít"be supposed to",
  • vol (volitive) the event is understood as "wanted/intended" ... chtít"want", hodlat "intent",
  • poss (possibilitive) the event is understood as "possible" ... moci, dát se "can, be possible",
  • perm (permissive) the event is understood as "permitted" ... smět "be allowed to",
  • fac (facultative) the event is understood as "an ability (to do sth)" ... umět, dovést "can, be able, ?could",
  • decl (declarative) basic (unmarked) modality.
Modal auxiliaries in UMR

UMR treats modal auxiliaries in a similar way as temporal ones, they inform the modal dependency annotation (and are not identified as separate events), thus He might/should go to school. is identified as a single event.

However, the class of English "true modal auxiliaries" is different - it comprises the following verbs, based on 4-3-2:

  • no modal verb, declarative sentences (also will, non-intentional be going to)
  • strong epistemic modals: must, have to, also should, ought to, need
  • weak epistemic modals: may, might, ??shall, ??would --> neutral affirmative

Comparing the two approaches, we can conclude that the modal verbs are treated in PDT and UMR (more-or-less) in the same way. The only exception represents

  • vol (volitive) modals ... chtít"want", hodlat "intent"

which are treated as semi-modals in UMR.

TODO Semi-modal verbs - separate events in UMR

According to Part 3-1-3-3. TAM categories, and Part 3-3-1-3. State verbs like want , need, and dread are categorized as semi-modal.

These verbs are annotated as separate events, they take an infinitive verb as a complement (i.e., an autosemantic verb is annotated as its child, :ARG1 in case of want in the following example). In addition, :modal-predicate relation should be annotated between the semi-modal and the complement.

(w/ want-01
    :ARG0 (p/ person
    	:refer-person 3rd
	:refer-number singular)
    :ARG1 (g/ go-01
    	:ARG1 p
        :ARG4 (s/ school)
        :aspect performance
	**:modal-predicate w)**
    :aspect state
    :modal-strength full-affirmative)

This :modal-predicate relation is applied also for other verbs with an infinitive complement, see the following example:

I saw him knock on the door.

(s1s/ see-01
	:ARG0 (s1p/ person
		:refer-person 1st
		:refer-number singular)
	:ARG1 (s1k/ knock-01
		:ARG0 (s1p2/ person
			:refer-person 3rd
			:refer-number singular)
		:ARG1 (s1d/ door
			:refer-number singular))
		:aspect performance
		**:modal-predicate s1s)**
	:aspect state
	:modal-strength full-affirmative)

(s/ sentence
	:temporal ((past-reference :contained s1s)
		 (s1s :overlap s1k)))

The following types are also considered as semimodals in some sources:

  • desideratives ('want to')
  • conatives ('try to')
  • optatives ('wish that'), and
  • frustratives ('fail to')

ŠZ: just one semantic concept consisting of two words (want to go) (?) - based on the test of negation, similar to Valency-changing operations (?), (https://github.com/umr4nlp/umr-guidelines/blob/master/guidelines.md#part-3-1-3-2-Valency-changing-operations)

!! ASI nechat jako pomocná podle PDT, když to může být language specific !!!
POZOR … pokud v PDT jako plnovýznamové, směr závislosti???
o PDT: něco přislíbit nemůže (tj. řídící infinitive, závislé modální ve fokusu)
o UMR: “hlavní” predikát non-finite She wants to go to school … jak vytipovat ???

Phasal aspectual meaning - one event with aspect attribute label

Phasal aspectual meanings such as

  • inchoative (denoting an aspect of a verb expressing the beginning of an action, typically one occurring of its own accord),
  • completive (adds a sense of completeness to a word or phrase (e.g. in the phrase break up, up is a completive)), and
  • continuative (relating to the durative aspect or a durative verb or verb form, i.e., that expresses action continuing unbroken for a period of time),

are never identified as separate events, even if they are expressed through independent words. Instead, they will simply inform the aspect attribute label of the event they modify [Part 3-3-1. Aspect] (https://github.com/umr4nlp/umr-guidelines/blob/master/guidelines.md#part-3-3-1-Aspect)

ce	əsi	mu-re
paper	this	black-change (Manipuri)
'This paper has become black.'
(h/ have-mod-91					(h/ have-mod-91
    :ARG1 (c/ ce 'paper')			    :ARG1 (p/ paper)
    	:mod (s/ əsi 'this')			        :mod (t/ this)
    :ARG2 (m/ mu 'black')			    :ARG2 (b/ black)
    :aspect Performance)			    :aspect Performance)

Possible UMR Aspect values at the default level of granularity include:
Activity - process that does not end,
Habitual - occurs/occurred usually or habitually,
State - unspecified type of state,
Endeavor - process that ends without reaching a result state,
Performance - process that ends and reaches a result state.

Further, Process value is used for event nominals.

Phasal verbs in English

The UMR Guidelines provides several examples with phasal verbs in Part 3 with sentence -level annotation. However, the document-level annotation is not described there :-((

According to Part 3-3-1-4 Aspect/Activity, thus the following sentences will have same sentence-level annotation (regardless the phase verbs):

  • [en] He is playing the violin. 3-3-1-4 (2)
  • [en] He started playing the violin. 3-3-1-4 (3a)
  • [en] He kept on playing the violin. 3-3-1-4 (3b)
He started playing the violin.
(p/ play-01
	:ARG0 (p2/ person
		:ref-person 3rd
		:ref-number Singular)
	:ARG2 (v/ violin)
	:aspect Activity
	:modstr FullAff)

The UMR Guidelines do not provide a list of English phasal verbs.
(AMR does not elaborate this phenomenon.)

Phasal verbs in Czech

Mluvnice 3 ... nic :-((

Čas a modalita v češtině ???

In PDT, phasal verbs are conceived as lexical words (= autosemantics), thus they are represented as separate nodes in t-trees.

Criteria for distinguishing phasal verbs in Czech NESČ, Fázové sloveso:
(a) phasal meaning
(b) phasal verb combines with infinitives (c) of imperfective verbs (no other forms!)

NESČ lists the following verbs (not necessarily exhaustive):
inceptives: jmout se, začínat - začít
terminatives: přestávat - přestat
continuatives: zůstávat - zůstat

VALLEX provides a bit broader set of verbs (class of phase verbs / phase of action, both limited to those with an infinitive complementation):
inceptives: jmout se, počínat - počít, začínat - začít, započít
terminatives: přestávat - přestat, ustávat - ustat
continuatives: zůstávat - zůstat

When these verbs combines with infinitives, they should NOT be identified as separate events (based on examples 3-3-1-4 (2) and 3-3-1-4 (3a), (3b)).
Instead, they will be represented at the document level -- HOW? The Guidelines does not describe their treatment (unless I missed something important)! -->

MOVE TO MODAL DEPENDENCY ANNOTATION

UMR … modal dependencies moved to document-level annotation … esp. possible-01, obligate-01 Part 3-1-3-3. TAM categories

  • [en] The boy can go., example 2 (2b)
(g/ go-01  
    :ARG0 (b/ boy
    	:ref-number Singular)
    :aspect State
    :modstr NeutAff)

(s0/sentence
  :modal (AUTH :NEUT s0g))

The presence of modal verbs (or other verbs or expressions) also informs the attribute of epistemic strength :modstrand :modpred. As for :modstr

  • no modal verb; certainly, be sure, definitely, necessarily
    -> :modstr FullAff / FullNeg
    (complete certainty that the event occurs / does not occur )
  • must/must have, have to, expect that, deduce; probably, likely; ...;
    -> :modstr PartAff or PartNeg
    (there is strong, but not definitive certainty that the event occurs / does not occur)
  • can, may, might, possibly, likely
    -> :modstr NeutAff or NeutNeg (there is neutral certainty that the event occurs / does not occur; event is expressed positively / negation of event is expressed)