Strong Generative Capacity in HPSG - PowerPoint PPT Presentation

1 / 12
About This Presentation
Title:

Strong Generative Capacity in HPSG

Description:

... and in writing grammars that predict what linguistic entities are well-formed. ... subset of FSs that will be regarded as possible well-formed linguistic entities. ... – PowerPoint PPT presentation

Number of Views:54
Avg rating:3.0/5.0
Slides: 13
Provided by: patrickja2
Category:

less

Transcript and Presenter's Notes

Title: Strong Generative Capacity in HPSG


1
Strong Generative Capacity in HPSG
  • SRL A language for the description of formal
    grammars

2
Definitions
  • Formal Grammar A mathematical idealization of
    the mental grammar. A scientific theory which
    predicts tat an utterance token will be judged
    grammatical if and only if it is a model of the
    grammar.
  • SGC The set of canonical representatives of the
    isomorphism classes of singly generated models of
    the grammar.
  • SRL A particular logic designed by King(1989) in
    order to talk about species and their features
    with precision, both in describing linguistic
    entities and in writing grammars that predict
    what linguistic entities are well-formed.

3
HPSG 87, 94
  • 87 The formalization was based on the
    intuitionistic feature structures to model
    partial information about linguistic entities.
  • 94 Sharp distinction between feature structures
    and feature structure descriptions
  • Feature structures Total mathematical
    idealizations of linguistic entities
  • Descriptions Pick out that subset of FSs that
    will be regarded as possible well-formed
    linguistic entities.
  • Gap Relationship between HPSG and its SGC.

4
SRL
  • Fundamental ontological assumptions of HPSG 1.
    closed world assumption (fixed set of disjoined
    ontological categories-species), 2. total
    well-typedness (different kinds of species have
    different kinds of identifiable parts (features)
    which in turn must be of certain kinds (species).

5
SRL
  • An SRL language is determined by 2 disjoined sets
    of nonlogical symbols (species and features) and
    a notion of appropriateness that relates them to
    each other.
  • A class of interpretations each of which provides
    meanings for the nonlogical symbols and for
    descriptions.

6
SRL
  • Signature SltS,F,Agt, A is a total function from
    SxF to Pow(S) called the appropriateness
    function.
  • Interpretation of S IltU,S,Fgt
  • Each species denotes a set of objects, each
    object is in the denotation of exactly one
    species and S assigns to each object the unique
    species to whose denotation the object belongs
  • F assigns to each feature the partial function it
    denotes

7
SRL
  • The appropriateness function encodes a strict
    relationship between the denotations of species
    and features if A(s,f)empty set then the domain
    of F(f) does not intersect with the denotation of
    s.
  • E.g. A(republic,MONARCH) empty set

8
Formal vs Informal Grammars
  • HPSG 94
  • Sort hierarchy and set of principles. Sorts that
    label terminal nodes are called maximal) in the
    sense of maximally informative-specific).
  • A sort inherits the feature declarations of its
    supersorts.
  • The finite set of sort symbols is partially
    ordered, with sort symbols corresponding to more
    inclusive types lower in the ordering.

9
Formal vs Informal Grammars
  • Maximal HPSG sorts SRL species in that each
    linguistic object is in the denotation of one and
    only one maximal sort.
  • A Nonmaximal HPSG sort is the disjunction of its
    maximal subsorts (e.g. sign word phrase
  • HPSG feature SRL feature
  • An SRL A can express the HPSG feature
    declarations for the maximal sorts
  • The S of an SRL grammar can express the sort
    hierarchy component of an HPSG grammar.

10
Types vs isomorphism classes
  • Replacement of types with isomorphism classes
  • A unique (abstr.) feature structure is
    associated with each distinct isomorphism class
    of feature structures.
  • How? We associate with each object ? in an
    interpretation I a unique abstract feature
    structure Abst1(?), namely the abstract feature
    structure associated with the isomorphism class
    of the FS determined by the ?.

11
SGC
  • In case the interpretation happens to be a model
    of the grammar -gt the SGC of a grammar could be
    identified with the objects in a model of a
    grammar.
  • Linguistic objects in HPSG feature structure
    descriptions
  • Each interpretation determines a unique set of
    abstract feature structures. If the
    interpretation is a model of a given grammar,
    then the set of abstract feature structures
    determined by the model is the SGC of the grammar.

12
Kinds of SGCs for a grammar
  • The set of abstract FSs which are ABST1(?) for
    some object ? in some interpretation I which is a
    model of the grammar.
  • The set of abstract FSs determined by some
    exhaustive model (it contains for any object ? in
    any model, an object ? which determines the same
    abstract FS.
  • The set of all singly generated models that each
    belongs to a one isomorphism class.
  • Summary The SGC of an HPSG grammar is the set of
    its models which are abstract FSs.
Write a Comment
User Comments (0)
About PowerShow.com