
View metadata, citation and similar papers at core.ac.uk brought to you by CORE 北星論集(短) 第 8 号(通巻第46号) March 2010 Dependency Structure vs.Phrase Structure: Two Analyses of English Determiners* Takafumi MAEKAWA Contents pendency structure notation used in WG. 1Introduction (1)Dependency structure 2 Determiners in WG 2.1 D depends on N 2.2 N depends on D 2.3 Mutual dependency 2.4Problems WG is a monostratal grammatical frame- 3 Determiners in HPSG work,and it dispenses entirely with mul- 3.1 Assumptions tiple levels of syntactic representations and 3.2 Application 4Conclusion transformations that mediate among them. Grammatical relations or functions are shown by explicit labels, such as ‘subj 1Introduction (ect)’and‘comp(lement)’. Dependency In this paper,we will be concerned with structure is a pure representation of rela- two types of syntactic representations. tions between head and dependent. The Phrase structure has been adopted as the head is shown as the tail of an arrow, basis for sentence structure by most syn- and the dependent is at the point.1 Note tactic theories.In phrase structures,indi- that the number of nodes is in a one!to! vidual words combine to form constitu- one correspondence to the number of ents. However, Word Grammar (hence- words in the sentence: (1)has three forth WG: Hudson 1984,1990,2003a, nodes and the sentence which it repre- 2007) does not acknowledge any unit sents has three words.Consider the corre- larger than a word,and employs depend- sponding phrase structure in(2). ency structure instead.In this framework, all relationships are word!based.Depend- (2)Phrase structure encies and phrases are alternative ways of representing relationships between words. Some believe that dependency structure and phrase structure are merely notational variants(Gaifman 1965;Robinson1970). There are,however,significant differences between the two.(1)is an example of de- Key words: Head!Driven Phrase Structure Grammar,Word Grammar,Determiner,Dependency ―51― 北星論集(短) 第 8 号(通巻第46号) There are seven nodes in this phrase (3)Something terrible has happened,but structure: S,two NPs,two Ns,VP and Idon’t know what. V.Note that the dependency structure in (1)has just three nodes.This means that (4)*Idon’t know what is going to hap- dependency structures are simpler than pen,but I do know happened already. phrase structures. Furthermore, Hudson (2007:118)states that‘each word that Secondly,predicates such as wonder and has at least one dependent is the head of sure require a wh!word(or whether or if ) a phrase which consists of that word plus as its complement.In(5)and(6),cited (the phrases of)all its dependents’,mean- from Hudson(2003b: 633),this require- ing that phrases are implicit in the de- ment is satisfied by who and what,not pendency structure. If phrase structures come and happened . In these sentences can be derived from dependency struc- what is selected by the higher verb is the tures,phrases are redundant.It looks as wh!word,so the verb must depend on it though dependency structure should be in the subordinate clause. chosen as a basis for syntactic representa- tion rather than phrase structure. (5)I wonder (who)came.* Another difference between depend- (6)I’m not sure (what)happened.* ency structure and phrase structure is that the former can represent mutual de- What emerges from these pieces of evi- pendency but the latter cannot.The WG dence is that what and happened depend analysis of extraction of wh!words assumes on each other.In the framework of WG, mutual dependency between the wh!word the dependency structure may be either and the verbal head.Let us take the de- of the two diagrams in(7). pendency structure of a wh!question What happened? for example. The wh!pronoun (7) what is the subject of the verb happened , and a verb’s subject is one of its depend- ents.This means that what must depend on happened .However,there is also evi- dence that the verb depends on the wh! word.Hudson(1990:361!382;2003b; 2004) argues that the verb is a comple- ment of the wh!pronoun and thus depends on it.Firstly,the wh!pronoun can occur without the verb as in(3),but(4)shows that the verb cannot appear without the Thus,wh!interrogatives may involve a mu- wh!pronoun. These examples are from tual dependency(Hudson 1990:361!382, Hudson(2004:30). 2003b,2007:142).WG has analysed wh! relatives,that!relatives and free relatives ―52― Dependency Structure vs.Phrase Structure: Two Analyses of English Determiners in this fashion(Hudson 1990:383ff,2003 2.1 D depends on N a).We will also see that WG treats the Hudson(2004)argues that D depends on relation between determiners and nouns in N mainly based on Van Langendonck’s terms of mutual dependency. Such an (1994) evidence. According to Van Lan- analysis where two words are interdepend- gendonck (1994:250), the head of NP ent is not possible in phrase structure adjuncts is the noun rather than the de- frameworks. terminer.He points out that the meaning It seems, then, that we have good of the noun decides if the NP adjunct is reason for adopting dependency structure possible.The eligible nouns refer to times rather than phrase structure as the basis (8),places(9)and manners(10).2 for syntactic representation. The goal of this paper is to compare (8)I overslept each morning. Hudson’s(2004)dependency!based account (9)Put it this side of the line. of English determiners in WG with Van (10)It’sbesttodoitmy way. Eynde’s(2005,2006) phrase!based ac- count in Head!Driven Phrase Structure (11) is not acceptable because the noun Grammar(henceforth HPSG: Pollard and does not belong to any of the above se- Sag 1987,1994).We will show that in an mantic categories. empirically adequate analysis of English de- terminers the role of head must be disso- (11)*I overslept each house. ciated from the role of selector.We will argue that this conclusion leads to ruling Whether or not NP adjuncts are possible out the dependency account. is not only decided by the meaning of the The organisation of the paper is as fol- noun concerned.It is also decided by the lows.In section 2,we consider how WG noun itself.Consider(12). deals with the determiners.Section 3 then look at the HPSG analysis of determiners (12)I did it the usual way /*manner. put forth by Van Eynde(2005,2006).In the final section,we offer some concluding (12)shows that way canbeusedasan remarks. adjunct although its synonym manner can- not. Similarly, the noun cannot be re- placed by a personal pronoun such as it. 2 Determiners in WG In this section we will consider two kinds (13)I saw him this morning. of data.In 2.1 we will look at evidence (14)*Isawhimit. which might show that determiners de- pend on nouns; in 2.2 we will look at The above facts show that whether or not evidence for the opposite conclusion.The a noun can be used in an NP adjunct de- examples in this section are cited from pends on the noun involved and on its Hudson(2004)unless otherwise indicated. meaning.Therefore,it is quite reasonable ―53― 北星論集(短) 第 8 号(通巻第46号) to say that the noun rather than the de- way as (17), where extraposition is terminer is the head of NP adjuncts. blocked by names of .The grammaticality Second, singular, countable common of this example shows that people,rather nouns need determiners, but others do than some, is a direct dependent of the not.This means that whether D is neces- verb. If the head of a phrase is, as sary or not is decided by N. Hudson (2004:12) defines, the word Third,common nouns allow no more which is associated with words outside than one determiner in English. This is that phrase, the head of some people quite similar to verbs and prepositions should be people,not some. which typically allow no more than one It seems,then,we can conclude that complement. D must depend on N. Finally, let us consider extraposition from NP. The NP people who have been 2.2 N depends on D waiting ten years is continuous in(15),but We have seen some evidence that D de- therelativeclauseisextraposedin(16). pends on N.In this subsection,however, we will look at some data that lead us to (15) People [who have been waiting ten the opposite conclusion: N depends on years] are still on the list. D.Hudson(2004)gives three kinds of evi- dence. (16)People are still on the list [who have Many European languages allow a been waiting ten years]. preposition and a definite article to fuse intoasinglewordform.Oneexampleis However, extraporision is not possible if French du,which is a fusion of a preposi- the antecedent noun(people,in this case) tion de and a definite article le. is deeply embedded(Hudson 2004:20). (19)du(= de le)village‘from the vil- (17)*Names of people are still on the list lage’ [who have been waiting ten years]. This fusion only occurs in a PP where the This means that extraposition is only al- article introduces an NP which is the com- lowed if the antecedent noun is a direct plement of the preposition. Let us com- dependent of the verb to which the rela- pare this with the infinitival clause le voir tive clause is attached(are,in this case). hier in(20),where it is impossible to fuse Now,look at the following data. de le into du. (18)Some people are still on the list [who (20)J’ai oubli! de le voir hier. have been waiting ten years]. I have forgotton of him to!see yesterday ‘I forgot to see him yesterday.’ If people depends on some in(18),extra- position should be prohibited in the same Assuming that all the definite articles can ―54― Dependency Structure vs.Phrase Structure: Two Analyses of English Determiners be used as object pronouns
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages12 Page
-
File Size-