GRAMMATICAL RELATION

Oleh Irmayani(20401107093)
Traditional grammars mark great use of the notions of subject and object (and also of the distinction between direct and indirect object). This largely based upon the formal distinctions of noun phrases within a sentence such as John gave Bill a book, where John is subject, Bill indirect object and a book direct object,and these are defined by the position of the noun phrases relative to the verb and to one another. In Latin these grammatical relation, as they have been called, are a marked by inflection – by the case ( in the traditional sense ) of the nouns, they subject being in the nominative, the direct object in the accusative and the indirect object.

These grammatical relation are also important when we consider the category of voice ( active and

passive) in many languages. For, if we compare John played the piano and The was played by John, it is

apparent that, which is active. It is the subject in the second. The passive.. While John is the subject in the

first, but appears after by in the second. Intuitively, and informally, what we want to say is that the object of

the active sentence becomes the subject of the passive, while the subject moves to the position after by or

becomes the ‘agent’.There are, however, some complications. In English we find that the inderect object

may become the subject of the passive, as in Bill was given a book by john, as may the inderect object- A

book was given to Bill by John (though this might perhaps be seen as the passive of John gave a book to

Bill not John gave Bill a book).

As long as the terms ‘deep subject’ and ‘deep object’ are used to deal solely with formal relations of

this kind no real problems arise. But we may well be tempted to see the deep subject as the ‘doer’ and the

deep object as the ‘sufferer’; some linguistic have used the terms ACTOR AND GOAL to make this

distinction. There are, however , difficulties if we attempt to define them in semantic terms. For it is by no

mans true that the subject of a transitive verb can always be seen as one who ‘does’ something. Verb of

this kind should deter us from attempting to define actor in semantic terms. But even with action verbs, it is

not clear that we can clearly establish what meant by actor.

In spite of the absence of any clear semantic definitions for these grammatical relation some scholars, those

who have advocated RELATION GRAMMAR, have argued that they are talked about deep ‘subject ‘

and ‘object’, this was only an informal description and restricted solely to English – and Chomsky actually

saw no reason to use these terms. It is only terms of subject and object that we can make any general,

universal statements about active and passive. For, whatever the apparent differences in the various

languages, in all cases the object of the active becomes the subject of the passive and the subject of the

active is removed elsewhere.

Most of the arguments in favor of relation grammar are of a technical and syntactic nature, ranging over

numerous languages, and cannot be followed up here. But, in general, it seems to be the case that such

notions as subject and object are useful in many languages. However, there are some languages which

appear to have a different system of grammatical relation. One of the most obvious fact about subject and

object in languages like English is that some verbs, those that are transitive, typically have both subjects and

objects, while others, the intransitive verbs, have subject only. In talking about the noun phrase with the

intransitive verbs as ‘subject’ we are, of course, identifying, in relation terms, with the subject of the

transitive verbs, and the justification is found in formal features of the language.

This is of interest to us here because the distinction involved sometimes corresponds to that of intransitive

and transitive in English. Thus the contracts between intransitive and transitive in English bounce in English is

found in the basic (non-causative) and causative forms of the verb meaning ‘jump’. French and many other

languages in a similar way use a verb meaning ‘do’ or ‘make’: English intransitive and transitive cook are

translated into french by curie and fare curie.

Some linguistic have suggested that the transitivity distinction can be dealt with in terms of causative, the

transitive being seen as the causative, the transitive being seen as the causative of a basic non-causative

form. Thus John rang the bell is interpreted as ‘John’ caused the bell to ring and, by and extension of this

idea, John killed Bill as ‘John caused Bill to die’. Bu there are objections to this. First, there is a difference

between this purely semantic analysis of English and the formal features of Tigris and French (though this

might not disturb the advocates of generative semantics. Secondly, languages have causative of transitive as

well as (basically) intransitive verbs. Moreover, there seems to be no obvious motivation for the choice of

the transitive as the basic non- causative form. Tigris does not, as we might have expected from the

arguments about English, treat the intransitive ‘break’ as basic and the transitive as causative; the intransitive

is, in fact, indicated by a form with the passive prefix – taster.

Furthermore. There are degrees of plausibility in the causative analysis of English verbs. A causative

analysis of march in The sergeant marched the recruits is more reasonable than a similar analysis of

(transitive) ring or kill. It is clear that the recruits actively performed the action of marching and that the

sergeant caused them to do so, but bells that are rung and people who are killed do not actively performed

the actions of ringing and dying under causation. Natural language comprehension efforts may use

techniques based on simple string recognition, or on deeper analysis of grammatical relation. The lexical

scanner will use the former method. Efforts to develop grammatical scanner have generally proved

unsatisfactory due to underlying complexity of human language. Because the domain of palliative care is

relatively well-defined, simple phrase recognition methods can give immediate improvements in search

results without the complexity of grammatical method. This approach is also consistent with experience

based on expert system construction, which has shown that it is more useful to encode domain – specific in

a limited, artificial computer form than to attempt to have the computer master a poorly- bounded domain.

The aim of work reported in this paper is to evaluate the extent to which proposed system ofgrammatical

relations reflect the kinds of deep linguistic knowledge required for semantic, representation, in particular

for deriving semantic.grammatical relation either produced by or ex-tracked from the output of wide-

converge syntactic parses are currently used as input to shallow semantic parsers, which identify semantic

relation that exist between predictors (typically verbs) and their dependents. Predicate- argument structure

identified in this way can then be used in tasks link information extraction and question answering.

However, wide- converge stochastic parsers are only rarely used in dialogue systems. Traditionally,

interpretation modules of dialogue systems utilize specialized parses and semantic interpreters. Unlike in

information retrieval and question answering tasks, the system often needs to be connected to a knowledge

base which represents the state of the world and must be able to convert user utterances into knowledge

base queries. In addition to identifying negation , quantification, tense and modality.

We formulated four principles for deep grammatical relation representation. Firstly, grammatical relations

should, whenever possible, reflect relations between the predictors correspond to the same role assignment..

For example, the deep grammatical relation in passive constructions should be the same as those in the

active equivalents. And the analysis of a control verb construction like John persuaded Mary to dance

should make it clear that there is a ‘subject’ grammatical relations from dance to Mary similar to that in the

implied sentence Mary danced. Secondly, a grammatical relation should, whenever possible, appear only if

there is a an explicit selection restriction link between the words. For example, in a raising verb construction

like John expect Mary to dance, there should be no grammatical relation from the raising verb expected to

its object Mary. Also, where a preposition functions strictly as a syntactic role marker, as in the

construction John relies on Mary, it should have no place in the grammatical relation analysis. Now the two

analyzes are formally distinct:

a. the first is rooted at predicate in a closed path and the second at lit;

b. the definite external argument the bulb takes scope over the modifier lit in the first but over in a closed

path in the second.

The shared task data set contains numerous passive participles, most of which can be classified into the

following four groups depending on how the participle is used: (a) complement of passive auxiliary, (b)

complement of raising verb, (c) nominal post modifier, (d) nominal Pre-modifier. In all these case, our

system for deep grammatical relation annotation requires: that there is a relation from the passive participle

to the deep object and that this relation be the same as in the corresponding active declarative construction,

so that predicate-argument structure can be straightforwardly derived.

The conclusion of these, we have proposed a set of principles for developing a grammatical relation

annotation system for use with both shallow and deep semantic interpretation systems, in particular a tutorial

dialogue system.