Distribution

From UNL Wiki
Revision as of 11:25, 24 March 2010 by Martins (Talk | contribs)
Jump to: navigation, search

Distribution (or word order) refers to the study of the order of the syntactic constituents of a language.

Contents

Natural Language

In the UNLarium framework, the distribution of a form is informed in the grammar, if general, or in the dictionary, in case of compounds and modifiers (articles, determiners, adjectives and adverbs) that do not follow a regular pattern.

In English, for instance, articles are always premodifiers. Therefore, distribution of articles must not be informed in the dictionary, but stated through a rule in the grammar. The same applies to determiners (such as "this") and ordinary adjectives (such as "beautiful"), whose behaviour is default: adjectives and determiners are normally premodifiers. Only exceptions to the general rule - free order adjectives such as "possible": "it is the only solution possible" or "it is the only possible solution" - must be treated in the dictionary. However, adverbs, in English, may be premodifiers or postmodifiers, and it is quite difficult to predict their behaviour. Therefore, distribution of adverbs must be informed in the dictionary, and not in the grammar.

Distribution is also to be informed in the dictionary in case of compounds such as "bring home the bacon" which are generated from a base form ("bring").

Representing distribution in the dictionary

In the dictionary, distribution is informed by assigning the following features to the base form and its constituents (in case of compounds):

Examples

  • Order
    • very = BEF - In English, the intensifier "very" is a premodifier: He is very rich (He is rich very)
    • well = AFT - In English, the adverb of manner "well" is a postmodifier: He speaks well (He well speaks)
    • yesterday = FRE - In English, the adverb of time "yesterday" may come either before or after the modified: Now I go or I go now.
  • Adjacency
    • the = FAR (In English, the article "the" has no precedence over other modifiers: the small round black leather handbag (small the round black leather handbag).
    • after (in "look after") = IMM (In English, the preposition "after" must come right after the base form "come" in order to form the compound "look after": We look after them (We look them after)
    • down (in "put down") = NEA (In English, the adverb "down" must come right after the base form "put" in order to form the compound "put down", except for the complement: Put down that (Put down it)

Observations

Middle position is used only for words to be inserted inside others (i.e., between the prefix and the root, or the root and the suffix).
Adverbs coming between auxiliaries and verbs must be defined as premodifiers.
Order and adjacency may be combined to express specific distributions
BEF,IMM means that the word occurs right before the modified (as with English intensifiers)

Representing distribution in the grammar

In the grammar, distribution is defined through S-rules in the following format:

<SYNTACTIC ROLE>(<ORDER>,<ADJACENCY>);

Where:

  • <SYNTACTIC ROLE> is the syntactic role (VA, VC, VS, VH, etc) of the constituent in relation to the head; and
  • <ORDER> is the position of the constituent in relation to the head. It may assume one of the following values:
    • FNT (in the beginning of the sentence)
    • END (in the end of the sentence)
    • BEF or << (to the left of the head and before a blank space)
    • AFT or >> (to the right of the head and after a blank space)
    • > (immediately to the right of the head, i.e., without any blank space)
    • < (immediately to the left of the head, i.e., without any blank space)
  • <ADJACENCY> is the precedence of the constituent in relation to other constituents of the same phrase. It may assume one of the following values:
    • IMM (immediately: right after or right before the head)
    • NEA (near: precedence over other constituents except IMM)
    • FAR (distant: no precedence over other constituents)

Examples

VS(<<,IMM);
the specifier must be generated to the left of the verb before a blank space with precedence over any other constituent
VC(>>,FAR);
the complement must be generated to the right of the verb after a blank space without any precedence over other constituents

Observations

Order and adjacency may be represented in different rules
VS(<<); (the specifier must be generated to the left of the verb before a blank space)
VS(IMM); (the specifier must be generated with precedence over any other constituent)
Complex distribution
A single distribution rule may contain several distribution operations:
  • VS(<<)VS(IMM); (the same as "VS(<<,IMM);")
  • VS(<<)VC(>>); (the specifier will be generated to the left and the complement to the right of the head)
  • VS(FAR)VC(FAR); (both the specifier and the complement of the verb have no precedence over other constituents)
Conditional distribution
Conditional case-marking may be stated by defining the left side of the s-rule and coindexing it to the right side:
  • VC(>>); (unconditional distribution: the complement will be always generated to the right of the verb);
  • VC(PPR):=VC(<<); (conditional distribution: the complement will be generated to the left of the verb if a personal pronoun (PPR);
Adjacency
Adjacency must be informed when two constituents are to be generated in the same direction (otherwise, the system will simply follow the order of application of rules defined in the grammar)
  • VC(>>)VA(>>,FAR); (or "VC(>>)VA(>>)VA(FAR);", i.e., the complement comes nearer the head than the adjunct)

Adjacency states a gradient of proximity and should be assigned only to differentiate the priority of generation

  • VC(>>,NEA)VA(>>); or VC(>>)VA(>>,FAR); but there's no need for VC(>>,IMM)VA(>>,FAR);
  • VS(>>,IMM)VC(>>,NEA)VA(>>); or VS(>>,IMM)VC(>>)VA(>>,FAR); but there's no need for VS(>>,IMM)VC(>>,NEA)VA(>>,FAR);

Adjacency is limited to three values (IMM, NEA, FAR) because of the binary nature of branching in the X-bar approach. More complex structures should be reorganized as intermediary projections and only then related one another. See projection for further information.

Reordering
Reordering can be done in two different ways:
  • By Ph-rules, if the process involves neighbour items and affects only the surface structure of the phrase;
  • By attribute change (i.e., deleting and adding distribution features), such as in "VC(->>,<<);" (i.e.,delete the "after" attribute and add the "before" attribute)
The symbol ^ is used for negation and to control infinite recursion
  • VC(^>>):=VC(>>); (assign the "after" attribute to the complement of the verb if it does not have it yet)

UNL

Word order is not informed in UNL.

Software