• The goal of the syntactic analysis of a language is to
formulate a grammar which will specify the sentences in the data in terms of a set of rules.
• In this case, the analyst makes explicit the structure
and express generalizations about them.
• Two different types of rules will be presented :
Phrase structure rules and relational- dependency rules
• The rules of Grammar specify the way the from classes
than may combine in the language. • A distinction can be drawn between lexical form classes which are the lexical categories e.g., noun, verb, adjective, adpostion. • and phrasal form classes that are constituents like noun phrase, prepositional phrase and verb phrase, specified by the rules of the grammar. • The elements in the lexical form classes are stored in the lexicon.
• The lexicon complements the grammar in important
ways. Ch 6 Phrase-structure rules • The relations among nodes in this tree can be expressed by the following rules. • (‘XP’ is an alternative notation for ‘X″’.) • ‘SPEC’, ‘MOD’ and ‘CMPL’ are functions, not form classes therefore a more accurate rendering in terms of form classes alone is given in (5.2), in which ‘WP’, ‘YP’ ‘ZP’ stand for maximal projections headed by ‘W’, ‘Y’ and ‘Z’, which are variables for lexical categories. (5.2) a. XP = (W(P)) + X′ b. X′ = (YP) + X′ c. X′ = (ZP) + X • The basic structure of the English noun phrase can be expressed as in (5.3); these rules are for ‘simple’ NPs. • NPs without sentential elements such as relative clauses. a. NP = (DET) + N′ (3.5) b. N′ = (AdjP) + N′ c. N′ = N + (PP) • These three statements are PS-rules. The most commonly used notation, however, • employs an arrow (‘→’) (5.4) a. NP → (DET) + N′ b. N′ → QP + N′ [optional] c. N′ → AdjP + N′ [optional] d. N′ → N + (PP) • The English PS-rules specify the structure of the English NP the three long proofs of the theorem. • First, the parenthesis ( )around DET and PP indicate that these constituents are optional • Second, the PS-rules expanding N′ in (5.4b) and (5.4c) are optional. • It is preferable to say that these rules are optional rather than AdjP and QP are optional (i.e. N′ → (QP) + N′, N′ → (AdjP) + N′), • since a literal interpretation of the latter could lead to an NP with three non-branching N′s; the minimal three-level structure can be specified by (5.4a) and (5.4d) alone. • Hence they are the only obligatory rules in (5.4), and if (5.4b) or (5.4c) apply, a QP or AdjP must occur in the NP. • Third, the plus (+) signals that there is a fixed order among the form classes specified in the rule, • i.e., DET precedes N′, AdjP precedes N′, QP precedes N′, and N precedes PP. • It should also be noted that rule (5.4c) can be repeated, in order to permit multiple adjectival modifiers as in the NP the beautiful expensive red scarf. Figure 4.11 • The rules in (5.1)–(5.4) are formulated in terms of the X-bar schema, but PS-rules could also be formulated to specify the non-X-bar phrase-structure. • The development of the X-bar schema came many years after the introduction of PS-rules. • Each level of X-bar structure in the tree is represented by a rule. • The constituent on the left side of the arrow is a mother node, dominates the daughter • node(s) on the righthand side of the rule. • Thus in (5.4a), ‘NP’ is a mother node and immediately dominates DET and N′, its daughters. Etc. • The correspondence between levels of structure in the tree and PS-rules can be seen in Figure 5.2 • Each level of structure in a tree, that is each mother node and its daughters, is represented by a PS-rule; in other words, each PS-rule specifies a local subtree.
• There are two ways to interpret the correspondence
between constituent structure and PS-rules.
• One way is to construe the rules as instructions for
drawing a tree . In this sense, they are often referred to as ‘rewrite rules’. • The other way to interpret PS-rules is as constraints on possible tree structures or node admissibility conditions. • PS-rules is used to determine whether the structure in question is possible in the language. • As an illustration of how this works, NP structure in several languages will be compared. • The PS-rules for a simple Indonesian NP like guru muda itu [teacher young that] ‘that young teacher’ are given in (5.5). • (5.5) PS-rules for simple Indonesian NP a. NP → N′ + (DET) b. N′ → N′ + AdjP [optional] c. N′ → N • As in English, determiners and adjectival modifiers are optional. • Five NP structures are given in Figure 5.3. • (5.5b). Which one is a possible NP structure for English? • Based on the PS-rules in (5.4), only (a) could be a possible English NP. Structures (b) and (d) violate rule (5.4b), while structures (c) and (d) violate rule (5.4a). • Structures in Figure 5.3, the PS-rules in (5.4) for English and (5.5) for Indonesian admit only one of them as potential grammatical structure in each language. • This is what is meant by treating PS-rules as constraints on possible structures. • Are structures (b) and (c) grammatical NP structures in some languages? • Structure (b) represents the structure of Spanish NPs • like el gato gordo [the cat fat] ‘the fat cat’
• It was mentioned earlier that rules like ‘NP → (DET) +
N′’ in (5.4a) express two different pieces of information: • (1) NP immediately dominates DET and N′. • )2) DET precedes N′. • A reformulation of the Indonesian PS-rules in (5.5) along this line is given in (5.6). (5.6) PS-rules for simple Indonesian NP (revised) a. Immediate dominance 1. NP → N′, (DET) 2. N′ → N′, AdjP [optional] 3. N′ → N b. Linear precedence Head-initial The rules in (5.6a) are called immediate-dominance rules [ID-rules]. • They differ from the earlier PS-rules formally only in the replacement of the plus (+) with a comma (,). • The ordering of the elements is provided by the linear precedence rule [LP-rule]. • When PS-rules are broken down into distinct ID- and LP-rules, certain generalizations can be captured that would otherwise be missed or difficult to express. • The single set of ID-rules in (5.6a) describes the first four of the tree structures in Figure 5.3; the structures differ only in terms of LP-rules. • So, instead of needing four sets of PS-rules, i.e. those in (5.4) for English, and other Ls, only one set of ID-rules, those in (5.6a), • and four language-specific LP-rules would be required for the specification of the (a)–(d) NP structures in Figure 5.3 • This format also permits the simplification of particular grammars. • In English, it is possible to have an adverb within the VP in two positions, either before • or after the V′, as in The lawyer quickly objected versus The lawyer objected quickly. • In the traditional format, two rules would be required to license these structures, V′ → AdvP + V′ and V → V′ + AdvP, whereas in the ID/LP format, only one would be needed, V → AdvP, V′. • While the formulation of ID-rules seems relatively straightforward in terms of the X- bar schema, the formulation of LP-rules can be quite complex. • Consider the English NP the three long proofs of the theorem in the book in Figure 5.4, which contains an adjunct PP. • To account for this structure, it is necessary to add an additional ID-rule to the set of ID-rules based on (5.4); the revised set of ID-rules is given in (5.7). • The new rule is (5.7b), which licenses the adjunct PP in the book. • The English NP is not consistently head-initial; • it is head-initial in the structures specified by (5.7b) and (5.7e), but is not in those permitted by (5.7a), (5.7c) and (5.7d). • It appears that two LP rules are required. • The first is for the head-initial structures in (5.7b) and (5.7e); in both cases the non-head daughter is a PP, and consequently the rule is ‘Head > PP’, • Where ‘>’ means ‘linearly precedes’. • The situation is more complicated in the head- final structures in (5.7a), (5.7c) and (5.7d). • In both cases the pre-head constituent is a kind of modifying element, a determiner in (a), a QP in (c) and an AdjP in (d). • But ‘modifying element’ is neither a form class nor a type of constituent. • therefore if the categories used in LP-rules are to be the same as those in the ID-rules, as they should, then this notion cannot be used in an LP-rule • There is no syntactic feature which unifies them, either, because determiner and quantifier are not among the major categories defined in terms of [±N] and [±V]. • The obvious but unsatisfactory solution is to have a disjunctive characterization of the pre- head constituent, • i.e. DET, QP or AdjP; this can be represented as ‘{DET, QP, AdjP} > Head’ • ‘{X, Y}’ means ‘X or Y’. The two LP-rules can be combined into a single rule. (5.8) LP-rule for simple English NP {DET, QP, AdjP} > Head > PP • This LP-rule and the ID-rules in (5.7) can together specify the English NPs like the ones in Figures 5.2 and 5.4. • The discussion of PS-rules thus far has been limited to NP structure, in the interest of introducing a number of basic concepts.